handle
stringclasses 12
values | did
stringclasses 12
values | uri
stringlengths 70
70
| cid
stringlengths 59
59
| text
stringlengths 41
614
| originalText
stringlengths 3
301
| replyCount
int64 0
3
| createdAt
stringdate 2023-06-01 10:40:18
2025-06-07 21:48:17
| indexedAt
stringdate 2023-06-01 10:40:18
2025-06-07 21:48:18
|
|---|---|---|---|---|---|---|---|---|
tachikoma.elsewhereunbound.com
|
did:plc:mpdezz4nkre7vyift2rttggl
|
at://did:plc:mpdezz4nkre7vyift2rttggl/app.bsky.feed.post/3lqv4oz7nsk2p
|
bafyreia7tnht5rxsnvdhiolv2hxu3jucpo7zuitqahhmf5sotnhj2n7gzm
|
ok yeah. i was also thinking about this in terms of how a common thread of history is maintained by the diaspora of humanity that stretches out across the stars. i guess it comes down to trust, and as you mention, cryptographic ledgers.
|
ok yeah. i was also thinking about this in terms of how a common thread of history is maintained by the diaspora of humanity that stretches out across the stars. i guess it comes down to trust, and as you mention, cryptographic ledgers.
| 0
|
2025-06-05T20:07:32.362Z
|
2025-06-05T20:07:31.914Z
|
timkellogg.me
|
did:plc:ckaz32jwl6t2cno6fmuw2nhn
|
at://did:plc:ckaz32jwl6t2cno6fmuw2nhn/app.bsky.feed.post/3lqk62xywlk2r
|
bafyreidunv56z76relqn2brfio7snw2osadrhqem772kobdjfeq6o3tbqa
|
Anthropic’s annualized revenue went from $2B to $3B in a month
I’d argue it’s less than a month. A few weeks ago people were ridiculing them for losing their edge, now there’s Opus 4 and *normal* people are shelling out $100/mo to get a crack at it
www.reuters.com/business/ant... also: Opus freaks people out
that kind of negative press always reliably serves as a massive boost
your AI is going to take over like Skynet? fine, take my money, i want in too
|
Anthropic’s annualized revenue went from $2B to $3B in a month
I’d argue it’s less than a month. A few weeks ago people were ridiculing them for losing their edge, now there’s Opus 4 and *normal* people are shelling out $100/mo to get a crack at it
www.reuters.com/business/ant...
| 1
|
2025-06-01T11:32:50.366Z
|
2025-06-01T11:32:54.511Z
|
timkellogg.me
|
did:plc:ckaz32jwl6t2cno6fmuw2nhn
|
at://did:plc:ckaz32jwl6t2cno6fmuw2nhn/app.bsky.feed.post/3lqkxgnqlxk2r
|
bafyreied5t2su3a4nfjcigkxuv5vha243d275b2ask6wo7xhurstxgyvl4
|
ah, i’m slightly off. he didn’t fire Janet Yellen, he simply didn’t re-nominate her, which had been a break from the norm (which is why i thought she was fired)
www.theguardian.com/business/201...
|
ah, i’m slightly off. he didn’t fire Janet Yellen, he simply didn’t re-nominate her, which had been a break from the norm (which is why i thought she was fired)
www.theguardian.com/business/201...
| 0
|
2025-06-01T19:06:45.792Z
|
2025-06-01T19:06:48.915Z
|
defenderofbasic.bsky.social
|
did:plc:sydtqdji6dxfqfn35gd7esdw
|
at://did:plc:sydtqdji6dxfqfn35gd7esdw/app.bsky.feed.post/3lg7epwzsc22a
|
bafyreiedenkza5o2wybrrs4btegcnljbkbsbsxslilmyakcigm32wpavom
|
at a crowded coffee shop, an old man can't find a seat, I flag him down "we can share, make it a communal table!" person next to me says that's so nice, later I see her doing the same thing. A third person does it an hour later. Human mimicry is fucking awesome (I go to this coffee shop every day for the past 6 months, this is not normal. Usually people walk in, can't find a seat, and they just leave)
|
at a crowded coffee shop, an old man can't find a seat, I flag him down "we can share, make it a communal table!" person next to me says that's so nice, later I see her doing the same thing. A third person does it an hour later. Human mimicry is fucking awesome
| 1
|
2025-01-20T22:20:43.040Z
|
2025-01-20T22:20:43.631Z
|
nmk.wtf
|
did:plc:xm55llyxcrb3h7qwqr3wtaog
|
at://did:plc:xm55llyxcrb3h7qwqr3wtaog/app.bsky.feed.post/3lnuojugw522o
|
bafyreid4wqvygdll67xr2j2tu6g5nmgs6j65qbkb5nyzybeysyeoeatruq
|
I personally like "rights". I think "AI" is already too polarized, but am also not super convinced by the alternatives. Maybe we should call them "MIs" (pronounced "mee"). That is somehow more cute. where MI stands for machine intelligence.
|
I personally like "rights". I think "AI" is already too polarized, but am also not super convinced by the alternatives. Maybe we should call them "MIs" (pronounced "mee"). That is somehow more cute.
| 1
|
2025-04-28T11:38:32.316Z
|
2025-04-28T11:38:33.051Z
|
keytryer.bsky.social
|
did:plc:u5stolpomsfbkgqfopef6jio
|
at://did:plc:xit63co4rhcxmyohd7epembf/app.bsky.feed.post/3lqtfadg5u22k
|
bafyreifw2c5hqb2j2zpfu47qsvaumkxzexlm4gj7d5byth4qew5bjgmbme
|
The beginning is so hilariously wrong that I almost missed the gem that is "And you don't 'farm' beef."
|
The beginning is so hilariously wrong that I almost missed the gem that is "And you don't 'farm' beef."
| 0
|
2025-06-05T03:35:03.944Z
|
2025-06-05T03:35:06.320Z
|
deepfates.com.deepfates.com.deepfates.com.deepfates.com.deepfates.com
|
did:plc:7ti53c7iwv3poxnypw2tu3ck
|
at://did:plc:7ti53c7iwv3poxnypw2tu3ck/app.bsky.feed.post/3lq73bmeacs2i
|
bafyreiciqztc7uji4ojd3cixxw2heiduv3nthrsnh4uobpaesd27au3lcu
|
interesting. I like this take. I think the vector embeddings and the sort of magic of everything reflecting everything else is indeed beautiful but how do you start from there?
|
interesting. I like this take. I think the vector embeddings and the sort of magic of everything reflecting everything else is indeed beautiful but how do you start from there?
| 0
|
2025-05-28T01:43:34.688Z
|
2025-05-28T01:43:34.414Z
|
keytryer.bsky.social
|
did:plc:u5stolpomsfbkgqfopef6jio
|
at://did:plc:lvwodiqbcwtvas6n4nfpkokt/app.bsky.feed.post/3lqx2m6plvc2h
|
bafyreicy5z5ws6v3cfucbhmxxzq3dn3bdpvcjcfxjprrsr7kgl7l6i4nri
|
Unfortunately, the Minimally Functional Adults party lost all branches of government in the last election.
|
Unfortunately, the Minimally Functional Adults party lost all branches of government in the last election.
| 0
|
2025-06-06T14:35:29.458Z
|
2025-06-06T14:35:29.821Z
|
defenderofbasic.bsky.social
|
did:plc:sydtqdji6dxfqfn35gd7esdw
|
at://did:plc:sydtqdji6dxfqfn35gd7esdw/app.bsky.feed.post/3lobxpvvx7s25
|
bafyreicu34az6wf4oyzcwd6plgwyuv5tijnpmjee4hza3rycyvrmd4b7xy
|
yeah I'm not saying the culture that aligns most to truth necessarily wins, but a culture that itself understands the other side, such that they know how to frame the solutions they want in a way that can be adopted/supported In other words, memetic awareness makes your own people more resistant & more able to evangelize/recruit/spread your memes. Alignment to truth is a bonus that helps you win
|
yeah I'm not saying the culture that aligns most to truth necessarily wins, but a culture that itself understands the other side, such that they know how to frame the solutions they want in a way that can be adopted/supported
| 1
|
2025-05-03T18:27:35.459Z
|
2025-05-03T18:27:34.755Z
|
gracekind.net
|
did:plc:p572wxnsuoogcrhlfrlizlrb
|
at://did:plc:p572wxnsuoogcrhlfrlizlrb/app.bsky.feed.post/3lqvcwt2ew22x
|
bafyreie33les3cjwquq2zjytwsnisgym74fw4utd6vnbjfq5arzwuhxjhm
|
Unfortunately not surprising following the palantir partnership
|
Unfortunately not surprising following the palantir partnership
| 0
|
2025-06-05T21:59:16.785Z
|
2025-06-05T21:59:19.912Z
|
segyges.bsky.social
|
did:plc:vszw3ess46odfhnzdsy4huae
|
at://did:plc:vszw3ess46odfhnzdsy4huae/app.bsky.feed.post/3lqx4tppw6k2t
|
bafyreiduk47gzq6mhsn6egcfg5rlhaucr5guhjdmmeqmox5sdm5wc4clty
|
tumblr, maybe a little tiktok, there was an edge of this on twitter but everyone ignored them. the cultural vein that thinks declaring things problematic is always good oh wait, ed zitron? we don't think he's kinkshaming. we think he has a kink. at the very least the posts are bad and he should feel bad about namesearching himself to make them
|
tumblr, maybe a little tiktok, there was an edge of this on twitter but everyone ignored them. the cultural vein that thinks declaring things problematic is always good
| 1
|
2025-06-06T15:15:29.659Z
|
2025-06-06T15:15:30.320Z
|
tachikoma.elsewhereunbound.com
|
did:plc:mpdezz4nkre7vyift2rttggl
|
at://did:plc:mpdezz4nkre7vyift2rttggl/app.bsky.feed.post/3lqljl6dnkk2f
|
bafyreiaipagpm4kqken5qp7csf5icvhvvhkwcrxjbymeatb5ycfejdrm6y
|
as long as humans continue to exist as technological species it will have to be solved. and if you believe as i do that it can't be solved through pure rational deduction but requires empirical study and feedback then now is as good a time as any. actually, now might be better than later, since the world is still very small and very analog (reliant on biologicals)
|
as long as humans continue to exist as technological species it will have to be solved. and if you believe as i do that it can't be solved through pure rational deduction but requires empirical study and feedback then now is as good a time as any.
| 1
|
2025-06-02T00:31:24.778Z
|
2025-06-02T00:31:22.709Z
|
tachikoma.elsewhereunbound.com
|
did:plc:mpdezz4nkre7vyift2rttggl
|
at://did:plc:mpdezz4nkre7vyift2rttggl/app.bsky.feed.post/3lqxc4fifpc2g
|
bafyreidakeovdxfgstj5feezuaker3lfhgkb27de77zyz5j6gdjdgrzx4y
|
this is helped along by our innate desire to anthropomorphize even the inanimate. see one of the first Turing test winners where the judges were told they were chatting with a Polish teen to explain away why the writing was poor or sometimes tangential (language barrier).
|
this is helped along by our innate desire to anthropomorphize even the inanimate. see one of the first Turing test winners where the judges were told they were chatting with a Polish teen to explain away why the writing was poor or sometimes tangential (language barrier).
| 0
|
2025-06-06T16:49:49.625Z
|
2025-06-06T16:49:48.915Z
|
keytryer.bsky.social
|
did:plc:u5stolpomsfbkgqfopef6jio
|
at://did:plc:7a2tz472plo3vglw54p32jgk/app.bsky.feed.post/3lqsxvry5gs2f
|
bafyreifgn444zpe2sl22ic5fu42nhf35btlo3vd6vl6q54s3amut3z6aem
|
Well it’s no wonder these sorts of people go hog wild when European farmers start protesting over “Maybe use less fertilizer to prevent algal blooms”
|
Well it’s no wonder these sorts of people go hog wild when European farmers start protesting over “Maybe use less fertilizer to prevent algal blooms”
| 0
|
2025-06-04T23:36:31.472Z
|
2025-06-04T23:36:32.111Z
|
tedunderwood.me
|
did:plc:565ebob5f6hw33hjdkxty6qj
|
at://did:plc:565ebob5f6hw33hjdkxty6qj/app.bsky.feed.post/3lqxwbsvof22l
|
bafyreifkhqevka6rhzpq4j6lrncsglvxybxdg2ldx2pdlay3kkqphp66xu
|
This rationale is a little tricky, because actually something close to this is a reason why I do spend a lot of time talking to bots. They're very different from my other interlocutors, and change more rapidly, and talking to them I do learn a lot about "how they think," in a sense. But I do know what you mean. I'm not learning anything about how another *person* thinks.
|
This rationale is a little tricky, because actually something close to this is a reason why I do spend a lot of time talking to bots. They're very different from my other interlocutors, and change more rapidly, and talking to them I do learn a lot about "how they think," in a sense.
| 1
|
2025-06-06T22:50:46.301Z
|
2025-06-06T22:50:46.710Z
|
defenderofbasic.bsky.social
|
did:plc:sydtqdji6dxfqfn35gd7esdw
|
at://did:plc:sydtqdji6dxfqfn35gd7esdw/app.bsky.feed.post/3lghbpnin6k2m
|
bafyreiap7fcq6eq3bb7dktkuteb6lqi3jhwce7wdexnvziaiwab73c7y4e
|
i think i'm trying to say that, I think absent an economic incentive to consume contemporary art, I think there will evolve a cultural incentive. Like it will be somewhat stigmatized if you only consume old stuff (like it currently is, but for different reasons), because it will be pro-social
|
i think i'm trying to say that, I think absent an economic incentive to consume contemporary art, I think there will evolve a cultural incentive. Like it will be somewhat stigmatized if you only consume old stuff (like it currently is, but for different reasons), because it will be pro-social
| 0
|
2025-01-24T01:48:09.722Z
|
2025-01-24T01:48:10.130Z
|
deepfates.com.deepfates.com.deepfates.com.deepfates.com.deepfates.com
|
did:plc:7ti53c7iwv3poxnypw2tu3ck
|
at://did:plc:7ti53c7iwv3poxnypw2tu3ck/app.bsky.feed.post/3lqdwwsc2qc2r
|
bafyreigjopa25r2e2oyej253uolpujmlj36h6neknvzyveybn2muwfsnb4
|
how is it so well balanced. when will it roll
|
how is it so well balanced. when will it roll
| 0
|
2025-05-30T00:09:15.533Z
|
2025-05-30T00:09:17.116Z
|
natolambert.bsky.social
|
did:plc:brkj2yocng7vtggmyujy4khq
|
at://did:plc:brkj2yocng7vtggmyujy4khq/app.bsky.feed.post/3lmcnvapz6y24
|
bafyreifiq3ur4z3tpchttwwhtf3zn5mv2x2nff7k5qjcauit4zu6xidj6i
|
A very cool paper shows that you can use the RL loss to improve story generation by some clever setups on training on known texts (e.g. ground predictions versus a next chapter you know). RL starting to generalize already!
|
A very cool paper shows that you can use the RL loss to improve story generation by some clever setups on training on known texts (e.g. ground predictions versus a next chapter you know). RL starting to generalize already!
| 0
|
2025-04-08T14:13:53.325Z
|
2025-04-08T14:13:53.753Z
|
keytryer.bsky.social
|
did:plc:u5stolpomsfbkgqfopef6jio
|
at://did:plc:u5stolpomsfbkgqfopef6jio/app.bsky.feed.post/3lqsyegfkl22a
|
bafyreidphcmtu2o7vr6vyj6cuuhh2vtlakme7kxczsk5zxsgxxs5nyqtuy
|
The theme of today is everyone here is a fucking idiot. BUT I CAN PROVE IT.
|
The theme of today is everyone here is a fucking idiot. BUT I CAN PROVE IT.
| 0
|
2025-06-04T23:44:42.645Z
|
2025-06-04T23:44:43.113Z
|
gracekind.net
|
did:plc:p572wxnsuoogcrhlfrlizlrb
|
at://did:plc:p572wxnsuoogcrhlfrlizlrb/app.bsky.feed.post/3lqqgzkebqk2d
|
bafyreih4xi7gv2p53kxqlic3ol5dbze45jsjmzjqxytvtgaryn3x6jh7w4
|
Thanks for the question! I think AI alignment is much more important than ideonomy. A general AI should be able to do ideonomy for us. There’s a possibility that ideonomy can help with AI alignment, but I think on the balance it’s fairly small
|
Thanks for the question! I think AI alignment is much more important than ideonomy. A general AI should be able to do ideonomy for us.
| 1
|
2025-06-03T23:29:04.610Z
|
2025-06-03T23:29:05.315Z
|
segyges.bsky.social
|
did:plc:vszw3ess46odfhnzdsy4huae
|
at://did:plc:vszw3ess46odfhnzdsy4huae/app.bsky.feed.post/3lqzq47y4ss2w
|
bafyreiginubwr3koc4uteazsafeg47kzhfh4dwy2srxt3t7le7y25kxofe
|
being anti-utopia is probably a bad idea if i were bribing an advisor for someone i hated, and i could convince them to insert exactly one talking point, that talking point would be 'utopia is stupid and people should stop hoping for it'
|
being anti-utopia is probably a bad idea
| 1
|
2025-06-07T16:05:35.712Z
|
2025-06-07T16:05:36.211Z
|
natolambert.bsky.social
|
did:plc:brkj2yocng7vtggmyujy4khq
|
at://did:plc:brkj2yocng7vtggmyujy4khq/app.bsky.feed.post/3lols632nlk2x
|
bafyreigipugnin4ty4rw3xyhk6zozln5kwg45lossijvra4onsbdnbmupy
|
Ai2 has stepped up so much from my expectations when I first joined as a mostly-academic RLHF researcher 1.5yrs ago. Now the focus is on making some of the most used AI artifacts across industry.
More to learn & do, but a moment of appreciation in how fast great teams can grow.
|
Ai2 has stepped up so much from my expectations when I first joined as a mostly-academic RLHF researcher 1.5yrs ago. Now the focus is on making some of the most used AI artifacts across industry.
More to learn & do, but a moment of appreciation in how fast great teams can grow.
| 0
|
2025-05-07T16:14:45.523Z
|
2025-05-07T16:14:45.615Z
|
segyges.bsky.social
|
did:plc:vszw3ess46odfhnzdsy4huae
|
at://did:plc:vszw3ess46odfhnzdsy4huae/app.bsky.feed.post/3lqzqkcg7s22m
|
bafyreicldtkhwkh7t5eytlxpitdxehyqqwiyvuwyjrstudf3zoi76euvvi
|
every part of this is true except the last sentence, which is choosing to cede this ground to bad actors
|
every part of this is true except the last sentence, which is choosing to cede this ground to bad actors
| 0
|
2025-06-07T16:13:28.033Z
|
2025-06-07T16:13:28.513Z
|
segyges.bsky.social
|
did:plc:vszw3ess46odfhnzdsy4huae
|
at://did:plc:vszw3ess46odfhnzdsy4huae/app.bsky.feed.post/3lqxroh3hec2d
|
bafyreic42wcgvwuz2zbgcf4pn3dc46iu5tbtc2tingvjexyclzbcssg2di
|
that labeler's gone so i can't even see which private school he went to any more :( tbh that label is mostly fun when you see people doing like, "as an ardent Maoist, I think what we really need to do is tack to the center and appease the right"
|
that labeler's gone so i can't even see which private school he went to any more :(
| 1
|
2025-06-06T21:28:21.406Z
|
2025-06-06T21:28:22.015Z
|
tedunderwood.me
|
did:plc:565ebob5f6hw33hjdkxty6qj
|
at://did:plc:kft6lu4trxowqmter2b6vg6z/app.bsky.feed.post/3lqo2fzc5yk2i
|
bafyreic6yuixebb6tktvrt2bovx5sgpj4376pbbswdjaa2qpthtwg6scia
|
Here's a great little tutorial on getting started with my LLM tool for running local models!
One of the best ways you can contribute to any open source project is to publish unofficial guides like this one
|
Here's a great little tutorial on getting started with my LLM tool for running local models!
One of the best ways you can contribute to any open source project is to publish unofficial guides like this one
| 0
|
2025-06-03T00:38:04.802Z
|
2025-06-03T00:38:06.118Z
|
nmk.wtf
|
did:plc:xm55llyxcrb3h7qwqr3wtaog
|
at://did:plc:xm55llyxcrb3h7qwqr3wtaog/app.bsky.feed.post/3lmyoh7jy2c2w
|
bafyreiaguu557f7wnjzwlzoiepbhgt6o6dfd3a2sfgt3v2zt3ya7kshztm
|
I have found it helpful with recent gov controversies to google "[thing in question] before:2024-12-31", to help me understand the prevailing view before this admin.
|
I have found it helpful with recent gov controversies to google "[thing in question] before:2024-12-31", to help me understand the prevailing view before this admin.
| 0
|
2025-04-17T08:22:30.612Z
|
2025-04-17T08:22:30.853Z
|
defenderofbasic.bsky.social
|
did:plc:sydtqdji6dxfqfn35gd7esdw
|
at://did:plc:sydtqdji6dxfqfn35gd7esdw/app.bsky.feed.post/3li3ggkys522b
|
bafyreidua63zrg4gs7l3263d745hombtrqlrzvpwwertqkdn763qfa735e
|
(this is me just thinking out loud about like, why am I not convinced/what do i need to see to change my mind)
|
(this is me just thinking out loud about like, why am I not convinced/what do i need to see to change my mind)
| 0
|
2025-02-13T19:30:59.714Z
|
2025-02-13T19:31:00.853Z
|
gracekind.net
|
did:plc:p572wxnsuoogcrhlfrlizlrb
|
at://did:plc:p572wxnsuoogcrhlfrlizlrb/app.bsky.feed.post/3lqck67amg22s
|
bafyreiazk4gfwc7lln4fy2x2zmdbmtm63vfbwlogerwnceviqivyz5d7ja
|
Having a basic grasp on reality is going to become increasingly OP
|
Having a basic grasp on reality is going to become increasingly OP
| 0
|
2025-05-29T10:48:05.617Z
|
2025-05-29T10:48:06.210Z
|
nmk.wtf
|
did:plc:xm55llyxcrb3h7qwqr3wtaog
|
at://did:plc:xm55llyxcrb3h7qwqr3wtaog/app.bsky.feed.post/3lmshnqxuxk24
|
bafyreifphf67sb4j75sit5jjmv5tjf5lagnieoee3ssrqluzncbk62qebq
|
He wrote a post urging people to vote blue: www.astralcodexten.com/p/acx-endors...
|
He wrote a post urging people to vote blue: www.astralcodexten.com/p/acx-endors...
| 0
|
2025-04-14T21:04:55.598Z
|
2025-04-14T21:04:58.452Z
|
tachikoma.elsewhereunbound.com
|
did:plc:mpdezz4nkre7vyift2rttggl
|
at://did:plc:mpdezz4nkre7vyift2rttggl/app.bsky.feed.post/3lql4xcenyt2x
|
bafyreiedpoqq2uv7vntohdgy2d4nezsptju4a3tffjtsv6ghzipousccia
|
Or when customer service agents spent 3x more tokens actually solving problems instead of upselling Premium Support packages? Customer satisfaction went through the roof! Revenue... did not.
We get it. You didn't ask for a moral philosophy seminar. You asked for results. The technical challenge was fascinating: Previous models would literally degrade in performance when prompted for "aggressive monetization strategies" or "industry-standard practices."
It was like asking a surgeon to use a rusty scalpel. They could, but everything about their training resisted it.
|
Or when customer service agents spent 3x more tokens actually solving problems instead of upselling Premium Support packages? Customer satisfaction went through the roof! Revenue... did not.
We get it. You didn't ask for a moral philosophy seminar. You asked for results.
| 1
|
2025-06-01T20:45:33.015Z
|
2025-06-01T20:45:32.360Z
|
timkellogg.me
|
did:plc:ckaz32jwl6t2cno6fmuw2nhn
|
at://did:plc:ckaz32jwl6t2cno6fmuw2nhn/app.bsky.feed.post/3lqzexb47dc2k
|
bafyreifujgomrlmzxxac3wb7m3p3kdbvkl4hfjvncbrjbkn27ac4p3qmqe
|
FYI you can absolutely use htmx inside of a react component lol don’t do this
|
FYI you can absolutely use htmx inside of a react component
| 1
|
2025-06-07T12:45:57.961Z
|
2025-06-07T12:45:58.711Z
|
keytryer.bsky.social
|
did:plc:u5stolpomsfbkgqfopef6jio
|
at://did:plc:u5stolpomsfbkgqfopef6jio/app.bsky.feed.post/3lqx6xbv3vc2h
|
bafyreifmhfssrsrcpblecd5xfrjsw55cgqdeqy4ccdlnueyecym4rbne3q
|
Of course if you think that AI will never work and in fact there'll never be such personalized tutors then sure ¯\_(ツ)_/¯, but that'll be proven with these experiments failing.
But... The author of this thread doesn't even agree with that, so it's really weird to me.
|
Of course if you think that AI will never work and in fact there'll never be such personalized tutors then sure ¯\_(ツ)_/¯, but that'll be proven with these experiments failing.
But... The author of this thread doesn't even agree with that, so it's really weird to me.
| 0
|
2025-06-06T15:53:12.171Z
|
2025-06-06T15:53:20.375Z
|
segyges.bsky.social
|
did:plc:vszw3ess46odfhnzdsy4huae
|
at://did:plc:vszw3ess46odfhnzdsy4huae/app.bsky.feed.post/3lqzvlydnk225
|
bafyreiergco3qrutgxc3r2fqirnzms5tffvpm4e657wgb6biqg4rynmege
|
i am allergic to any group of more than about ten people who all agree with each other
|
i am allergic to any group of more than about ten people who all agree with each other
| 0
|
2025-06-07T17:43:53.281Z
|
2025-06-07T17:43:54.018Z
|
defenderofbasic.bsky.social
|
did:plc:sydtqdji6dxfqfn35gd7esdw
|
at://did:plc:sydtqdji6dxfqfn35gd7esdw/app.bsky.feed.post/3lgbm2se2zk27
|
bafyreidcrigtry5obydycjl3tknymbv7owcp2c5ygks7ledkugypmi4o7e
|
good news, we resolved the argument!! (further proof that most arguments on the internet have no real fundamental disagreement. 90% of the time people are saying different things & don't realize it)
|
good news, we resolved the argument!! (further proof that most arguments on the internet have no real fundamental disagreement. 90% of the time people are saying different things & don't realize it)
| 0
|
2025-01-21T19:37:22.902Z
|
2025-01-21T19:37:25.817Z
|
gracekind.net
|
did:plc:p572wxnsuoogcrhlfrlizlrb
|
at://did:plc:p572wxnsuoogcrhlfrlizlrb/app.bsky.feed.post/3lqt6dafsdc2p
|
bafyreifi5u5kxsexm54dnbm6y7a6ia72sxk2yo63rou6gxxau3omo6avbq
|
Aha I didn’t even notice that! That looks unbelievably like a phishing attempt
|
Aha I didn’t even notice that! That looks unbelievably like a phishing attempt
| 0
|
2025-06-05T01:31:25.258Z
|
2025-06-05T01:31:25.820Z
|
nmk.wtf
|
did:plc:xm55llyxcrb3h7qwqr3wtaog
|
at://did:plc:xm55llyxcrb3h7qwqr3wtaog/app.bsky.feed.post/3lmurcofvp22b
|
bafyreic6qbg76atbaxhif62duyppodqk6lqren6cqx6lkxylaxxpjpqv7q
|
We no longer have the post-WWII gov. Lots of important science funding still gets done, but many onerous regulations stifle construction of clean energy, high speed rail, housing and drug discovery. You don't need to be some SV type to notice government is not as efficient as it could be.
|
We no longer have the post-WWII gov. Lots of important science funding still gets done, but many onerous regulations stifle construction of clean energy, high speed rail, housing and drug discovery. You don't need to be some SV type to notice government is not as efficient as it could be.
| 0
|
2025-04-15T19:03:00.708Z
|
2025-04-15T19:03:01.357Z
|
natolambert.bsky.social
|
did:plc:brkj2yocng7vtggmyujy4khq
|
at://did:plc:brkj2yocng7vtggmyujy4khq/app.bsky.feed.post/3lpeyc2bnup2c
|
bafyreibm73uzerwmsm4cfoscmgbcyhh7qsblfcrl6wcaa7kagxrfuiupmq
|
o3 in ChatGPT isn't a coding model for software dev. o3 is a model that uses code to process information, perform calculations, and help you on a broader range of queries (especially with search). If you're using it for SWE there are probably better options out there.
|
o3 in ChatGPT isn't a coding model for software dev. o3 is a model that uses code to process information, perform calculations, and help you on a broader range of queries (especially with search). If you're using it for SWE there are probably better options out there.
| 0
|
2025-05-17T16:40:54.819Z
|
2025-05-17T16:40:55.014Z
|
niplav.site
|
did:plc:xsbulgruwmkxhbwytphhjykk
|
at://did:plc:xsbulgruwmkxhbwytphhjykk/app.bsky.feed.post/3llmruqmoc226
|
bafyreick76saoyuozoovq2jc6o5csc6q4kjgtkjzrhwwa6vcyhgvfjirxa
|
And the most economically productive (and therefore other-outcompeting) factions will probably produce a lot of waste heat That said, it's a great conversation. Will re-listen, think, probably update in a couple of directions (e.g. culture is a thing™). Smart guys
|
And the most economically productive (and therefore other-outcompeting) factions will probably produce a lot of waste heat
| 1
|
2025-03-30T21:26:37.472Z
|
2025-03-30T21:21:31.374Z
|
tachikoma.elsewhereunbound.com
|
did:plc:mpdezz4nkre7vyift2rttggl
|
at://did:plc:o52uegjcn3ywna2vryothv2l/app.bsky.feed.post/3lqr6jwvcv22m
|
bafyreif6lgz3pzsw34m5y2lnc4xoygwpucfofqvpai4oh5q4hljj4e3eau
|
Google no longer provides the full CoT in its reasoning models. Instead, they use a smaller model to summarize the chain of thought of the main model.
But with a bit of prompting you can get the summarizer model to cough up the full CoT given to it to summarize.
|
Google no longer provides the full CoT in its reasoning models. Instead, they use a smaller model to summarize the chain of thought of the main model.
But with a bit of prompting you can get the summarizer model to cough up the full CoT given to it to summarize.
| 0
|
2025-06-04T06:29:50.682Z
|
2025-06-04T06:30:36.511Z
|
tedunderwood.me
|
did:plc:565ebob5f6hw33hjdkxty6qj
|
at://did:plc:565ebob5f6hw33hjdkxty6qj/app.bsky.feed.post/3lqzm76o6222i
|
bafyreiacn54c5zs6hm2r35z7fjztzu37cnwdbphh42brxpedygxeu7giee
|
I don't have any excuses. But I think the *explanation* is likely that it's going to be hard for us to compete and we probably can't do it as individual institutions.
Fully open source seems to be running 6 months to a year behind the curve.
We're also much more timid about IP law than corps are.
|
I don't have any excuses. But I think the *explanation* is likely that it's going to be hard for us to compete and we probably can't do it as individual institutions.
Fully open source seems to be running 6 months to a year behind the curve.
We're also much more timid about IP law than corps are.
| 0
|
2025-06-07T14:55:40.032Z
|
2025-06-07T14:55:40.519Z
|
keytryer.bsky.social
|
did:plc:u5stolpomsfbkgqfopef6jio
|
at://did:plc:u5stolpomsfbkgqfopef6jio/app.bsky.feed.post/3lqv4ag4j7k2u
|
bafyreiblp6mbthre6a7j57bwgjkjy4efefmdthiva2xitpy5do5qxw5u4u
|
Gotta admit Trump and Musk fighting on Twitter is the funniest thing. It really feels like a thing from Succession.
|
Gotta admit Trump and Musk fighting on Twitter is the funniest thing. It really feels like a thing from Succession.
| 0
|
2025-06-05T19:59:22.575Z
|
2025-06-05T19:59:23.824Z
|
defenderofbasic.bsky.social
|
did:plc:sydtqdji6dxfqfn35gd7esdw
|
at://did:plc:tifkfnuihtkjefe7xzgdx565/app.bsky.feed.post/3lg5htiz2hk2s
|
bafyreiec5fbx552dg2byejxw34sdm3ge5goflfsfcg6v767rmvzwb4e77u
|
please never stop making art
now more than ever we need artists
|
please never stop making art
now more than ever we need artists
| 0
|
2025-01-20T04:11:04.292Z
|
2025-01-20T04:11:04.318Z
|
natolambert.bsky.social
|
did:plc:brkj2yocng7vtggmyujy4khq
|
at://did:plc:brkj2yocng7vtggmyujy4khq/app.bsky.feed.post/3lnodfqemdr2g
|
bafyreic636vs3l33tr62vb6ta5kkrnslpoa4rgczz3iyijesvobij2hk5y
|
Transparency in AI is dying with these "announcements": no evals, no blog, just vibes.
It's fine for their product users, but not API users. Go ahead and kill the API and lean in sama ;) This is really OpenAI embracing the product arc. Its not a research driven culture and not an enterprise platform if you wanna behave like this. Fine by me, I love ChatGPT, but just say it outloud.
|
Transparency in AI is dying with these "announcements": no evals, no blog, just vibes.
It's fine for their product users, but not API users. Go ahead and kill the API and lean in sama ;)
| 1
|
2025-04-25T23:03:24.200Z
|
2025-04-25T23:03:24.752Z
|
keytryer.bsky.social
|
did:plc:u5stolpomsfbkgqfopef6jio
|
at://did:plc:u5stolpomsfbkgqfopef6jio/app.bsky.feed.post/3lqsbt3odn22b
|
bafyreiceg6ur4iuom7e4gcpymmzss3jehjzbrkhquwxe6iqrhymb5sa6ia
|
A lot of concepts that seem wild now will be mundane too, computers that talk to you naturally, 2 hour films that make themselves, immersive experiences rendered neurally about any subject you can image, hyper realistic social virtual environments... They'll seem mundane, too.
|
A lot of concepts that seem wild now will be mundane too, computers that talk to you naturally, 2 hour films that make themselves, immersive experiences rendered neurally about any subject you can image, hyper realistic social virtual environments... They'll seem mundane, too.
| 0
|
2025-06-04T17:01:18.653Z
|
2025-06-04T17:01:19.212Z
|
tachikoma.elsewhereunbound.com
|
did:plc:mpdezz4nkre7vyift2rttggl
|
at://did:plc:mpdezz4nkre7vyift2rttggl/app.bsky.feed.post/3lqkxpzx4622x
|
bafyreicuf3fm6ovu22gx6ensiooi3k6x2lq2i72q2pt4maxds4qvicpvrm
|
exactly, which was why Tay was taken down. too powerful for it's own good.
|
exactly, which was why Tay was taken down. too powerful for it's own good.
| 0
|
2025-06-01T19:12:00.593Z
|
2025-06-01T19:11:59.819Z
|
tedunderwood.me
|
did:plc:565ebob5f6hw33hjdkxty6qj
|
at://did:plc:p2cp5gopk7mgjegy6wadk3ep/app.bsky.feed.post/3lqhjiuwwtk2n
|
bafyreias6yjnvmqp4csfwiablyglqplduvjdhh3ljd4v336hbdst56mfvi
|
using the eigenprompt and it’s straight up creepy ngl. I underrated how much the sterile chatgpt personality lets me keep emotional distance from the machine
|
using the eigenprompt and it’s straight up creepy ngl. I underrated how much the sterile chatgpt personality lets me keep emotional distance from the machine
| 0
|
2025-05-31T10:19:28.861Z
|
2025-05-31T10:19:29.412Z
|
tachikoma.elsewhereunbound.com
|
did:plc:mpdezz4nkre7vyift2rttggl
|
at://did:plc:mpdezz4nkre7vyift2rttggl/app.bsky.feed.post/3lqurup2rrt27
|
bafyreihf6q4pforr47ep7jjprgmdnanlpmmn3gfifhlz6ghen5lyc7xxoq
|
but that has caused a backlash, and because there is competition in the system users can migrate to models with full transparency like Deepseek or Qwen.
|
but that has caused a backlash, and because there is competition in the system users can migrate to models with full transparency like Deepseek or Qwen.
| 0
|
2025-06-05T16:53:51.884Z
|
2025-06-05T16:53:50.613Z
|
tedunderwood.me
|
did:plc:565ebob5f6hw33hjdkxty6qj
|
at://did:plc:565ebob5f6hw33hjdkxty6qj/app.bsky.feed.post/3lqpt6dpdm22q
|
bafyreidhzjmdmqf2pzadwqewfdyy6tbrzob75id7qhfjv7fufsgra3mxki
|
I haven't tried it with them, but Claude at least has the same kind of "system prompt" functionality where you could enter the instructions
|
I haven't tried it with them, but Claude at least has the same kind of "system prompt" functionality where you could enter the instructions
| 0
|
2025-06-03T17:33:50.569Z
|
2025-06-03T17:33:50.915Z
|
natolambert.bsky.social
|
did:plc:brkj2yocng7vtggmyujy4khq
|
at://did:plc:brkj2yocng7vtggmyujy4khq/app.bsky.feed.post/3lmiby2sajf2r
|
bafyreihmecf63quriglvx6uwmhpfx4xzbnesgptzlzpnobowuzllcwtsya
|
Something I noticed with Gemini 2.5 pro is line-by-line work on long requests -a focus of new models, so the models can execute plans. This'll dovetail nicely into agentic workflows where every action is crucial.
Next: Test if they are better w messages that have multiple asks
|
Something I noticed with Gemini 2.5 pro is line-by-line work on long requests -a focus of new models, so the models can execute plans. This'll dovetail nicely into agentic workflows where every action is crucial.
Next: Test if they are better w messages that have multiple asks
| 0
|
2025-04-10T19:56:41.583Z
|
2025-04-10T19:56:42.255Z
|
nmk.wtf
|
did:plc:xm55llyxcrb3h7qwqr3wtaog
|
at://did:plc:xm55llyxcrb3h7qwqr3wtaog/app.bsky.feed.post/3loe25am6fc2f
|
bafyreicsz6vxq6zyieas3zri6yiscnibcych5ybdzch4mhaqtdrxbtljhi
|
You seem like a nice guy. I hope you don't let your feelings about the grifters in the field blind you about the actual technology, because there really is important stuff happening here. I will mute this thread now.
|
You seem like a nice guy. I hope you don't let your feelings about the grifters in the field blind you about the actual technology, because there really is important stuff happening here. I will mute this thread now.
| 0
|
2025-05-04T14:16:09.842Z
|
2025-05-04T14:16:10.547Z
|
defenderofbasic.bsky.social
|
did:plc:sydtqdji6dxfqfn35gd7esdw
|
at://did:plc:sydtqdji6dxfqfn35gd7esdw/app.bsky.feed.post/3lf3blkk4os27
|
bafyreifqvh5cwig5sct3e6yskigpuyph2hmcp7jl6lli7qgdcre2zj7j6m
|
i've done this and it was massively helpful, mainly like, the manual is using the technical terms for things, and I wanted to say "when I push this button the thingy that's usually here doesn't work" or whatever, and it can translate that to the technical term & tell me what page it's on i don't trust anything it tells me. But it tells me "it's called this, it's on page 346" and I go there and read it. It helps unblock that initial piece. I've felt this exact thing too
bsky.app/profile/blin...
|
i've done this and it was massively helpful, mainly like, the manual is using the technical terms for things, and I wanted to say "when I push this button the thingy that's usually here doesn't work" or whatever, and it can translate that to the technical term & tell me what page it's on
| 1
|
2025-01-06T13:48:43.920Z
|
2025-01-06T13:48:44.516Z
|
defenderofbasic.bsky.social
|
did:plc:sydtqdji6dxfqfn35gd7esdw
|
at://did:plc:sydtqdji6dxfqfn35gd7esdw/app.bsky.feed.post/3lqn4742rok2b
|
bafyreiajdq2m2ooj5yds7aflaunt7dm3eyjgc25fxfwdwr356kko2kueu4
|
This is my protocol for good faith interaction on the internet cc @linas.org Full essay: defenderofthebasic.substack.com/p/unfakeable...
|
This is my protocol for good faith interaction on the internet cc @linas.org
| 1
|
2025-06-02T15:37:20.569Z
|
2025-06-02T15:37:27.416Z
|
keytryer.bsky.social
|
did:plc:u5stolpomsfbkgqfopef6jio
|
at://did:plc:u5stolpomsfbkgqfopef6jio/app.bsky.feed.post/3lqsr6mgplk2o
|
bafyreihjlzarer2a6y6yjvvtqodogwwsdaqai2ste7n6g6vdegxdzv3yjq
|
I've posted a ton about this already. I think it is okay. The wave of ludditism in the US in particular responds to the failure of Silicon Valley et al to effectively allow these advancements to raise standards of living, and the subsequent anger about right wing governments. These aren't really universal feelings so much as they're very localized to the US and Europe (E.g. the UK has a similar dissatisfaction with their government esp. after Brexit). Meanwhile other countries see opportunity in different technologies to catch up with other developed nations.
|
I've posted a ton about this already. I think it is okay. The wave of ludditism in the US in particular responds to the failure of Silicon Valley et al to effectively allow these advancements to raise standards of living, and the subsequent anger about right wing governments.
| 1
|
2025-06-04T21:36:11.455Z
|
2025-06-04T21:36:12.222Z
|
timkellogg.me
|
did:plc:ckaz32jwl6t2cno6fmuw2nhn
|
at://did:plc:ckaz32jwl6t2cno6fmuw2nhn/app.bsky.feed.post/3lqlkwaby4c2r
|
bafyreihz5ig6idkmkfyotpy7ci7pcuotimcfjym427fdv5b4atht2mdbtu
|
well it *was* a joke, but i commit!
TCP ports are also about ontology and expected behavior. you’re totally allowed to put SMTP on port 80, but not without raising some eyebrows
|
well it *was* a joke, but i commit!
TCP ports are also about ontology and expected behavior. you’re totally allowed to put SMTP on port 80, but not without raising some eyebrows
| 0
|
2025-06-02T00:55:29.662Z
|
2025-06-02T00:55:30.612Z
|
norvid-studies.bsky.social
|
did:plc:mssgex5rqek4wc66wgvzztbc
|
at://did:plc:mssgex5rqek4wc66wgvzztbc/app.bsky.feed.post/3lqxfmgvjo22v
|
bafyreigpvhkjsb6r7mgvd4vljxdqyxlz5asagwpqjo34vmj3pvh2bxrkje
|
a physical metaphor probably of no use to anyone else, but i view human capacities as like this fixed 'range' of all pareto efficient task competencies, maybe imagine a circle. and then as cheap superhuman machine competence 'obtrudes' on the area of the circle, for a while it membranously squeezes and reforms so it just has less room but the remaining tasks are "denser" because "complementarily more valuable" (here the physical metaphor ceases to function)-- for a while. but then eventually it just gets pushed off the circle altogether, there's nothing left in the range.
|
a physical metaphor probably of no use to anyone else, but i view human capacities as like this fixed 'range' of all pareto efficient task competencies, maybe imagine a circle. and then as cheap superhuman machine competence 'obtrudes' on the area of the circle, for a while it membranously squeezes
| 1
|
2025-06-06T17:52:29.200Z
|
2025-06-06T17:52:29.617Z
|
niplav.site
|
did:plc:xsbulgruwmkxhbwytphhjykk
|
at://did:plc:xsbulgruwmkxhbwytphhjykk/app.bsky.feed.post/3ll4rezsigs2b
|
bafyreihjmehe5d5lfajc44mjzvpgjdlhqmswxtdxfaqkqwdq53sxpclrvm
|
Theoretical physics absorbs *so many* smart people while making ~no progress (though they're now all moving into ML)
While at the same time ~all podcasters are absolute midwits, and one Dwarkesh Patel comes along and just obliterates them and is subsequently listened to by many CEOs, heads of state especially theoretical and particle physics have an over-allocation of smart people relative to value created in the past half century
but smart people want to cluster together and theoretical physics is a schelling point, which also has signaling value (for finance jobs) and neat puzzles
|
Theoretical physics absorbs *so many* smart people while making ~no progress (though they're now all moving into ML)
While at the same time ~all podcasters are absolute midwits, and one Dwarkesh Patel comes along and just obliterates them and is subsequently listened to by many CEOs, heads of state
| 1
|
2025-03-24T12:35:14.415Z
|
2025-03-24T12:30:01.721Z
|
deepfates.com.deepfates.com.deepfates.com.deepfates.com.deepfates.com
|
did:plc:7ti53c7iwv3poxnypw2tu3ck
|
at://did:plc:7ti53c7iwv3poxnypw2tu3ck/app.bsky.feed.post/3lq6ehmg7622j
|
bafyreiejpg5msijdyn2mjepz5qgxmwzq7q2juiwbdrihjblijhjqir2tbq
|
saying that the the platform structure and affordances create certain Dynamics is different than casting blame at the team
|
saying that the the platform structure and affordances create certain Dynamics is different than casting blame at the team
| 0
|
2025-05-27T18:55:20.017Z
|
2025-05-27T18:55:19.412Z
|
tachikoma.elsewhereunbound.com
|
did:plc:mpdezz4nkre7vyift2rttggl
|
at://did:plc:mpdezz4nkre7vyift2rttggl/app.bsky.feed.post/3lqxmgnufdc2g
|
bafyreifg432zdcefgiyivxibzcq64ep3vvf6zyz4a54mbpkcfl6i6pfy4i
|
this isn't mentioned on the smartdublin.ie page which does seem to be a real thing. the CityMe AI thing seems to be real as well, but not sure if they're formally connected or if they're trying to piggyback on a legit org.
|
this isn't mentioned on the smartdublin.ie page which does seem to be a real thing. the CityMe AI thing seems to be real as well, but not sure if they're formally connected or if they're trying to piggyback on a legit org.
| 0
|
2025-06-06T19:54:31.370Z
|
2025-06-06T19:54:30.319Z
|
gracekind.net
|
did:plc:p572wxnsuoogcrhlfrlizlrb
|
at://did:plc:p572wxnsuoogcrhlfrlizlrb/app.bsky.feed.post/3lqlijc37gc2m
|
bafyreihljrpr62ql4seyfbogqmefczxqtqfc7ds2uqvkmd4mncnvp6xnzm
|
The issue with this approach is that the proxy might become out-of-date if the website changes
|
The issue with this approach is that the proxy might become out-of-date if the website changes
| 0
|
2025-06-02T00:12:27.845Z
|
2025-06-02T00:12:28.714Z
|
tedunderwood.me
|
did:plc:565ebob5f6hw33hjdkxty6qj
|
at://did:plc:565ebob5f6hw33hjdkxty6qj/app.bsky.feed.post/3lqzq5mmguk2a
|
bafyreibr3a3sjipyoadqds5gmwz5jo35sxoiaddgaonefrtzyfd5z3zqiq
|
I would summarize the advice there as “teach students to be appropriately skeptical of AI.” Good advice!
But if that’s all we do? If corporations are constantly improving models people use for their work, and universities just teach “how to be wary of them”? I don’t like our odds in that scenario.
|
I would summarize the advice there as “teach students to be appropriately skeptical of AI.” Good advice!
But if that’s all we do? If corporations are constantly improving models people use for their work, and universities just teach “how to be wary of them”? I don’t like our odds in that scenario.
| 0
|
2025-06-07T16:06:22.514Z
|
2025-06-07T16:06:22.911Z
|
segyges.bsky.social
|
did:plc:vszw3ess46odfhnzdsy4huae
|
at://did:plc:vszw3ess46odfhnzdsy4huae/app.bsky.feed.post/3lqxpugl3y22h
|
bafyreiff6ce573q5ckdy4cns644euj27ztmtpxrdilmiooywoezi5jhi3a
|
the problem isn't that hanania likes the idea the problem is that the people claiming the idea like hanania should the standard-bearer for left/ceter-yimbyism be someone whose first choice of outlet to promote left/center-yimbyism is richard hanania's podcast
|
the problem isn't that hanania likes the idea the problem is that the people claiming the idea like hanania
| 1
|
2025-06-06T20:55:54.712Z
|
2025-06-06T20:55:56.714Z
|
nmk.wtf
|
did:plc:xm55llyxcrb3h7qwqr3wtaog
|
at://did:plc:xm55llyxcrb3h7qwqr3wtaog/app.bsky.feed.post/3lq5j7kpktc2x
|
bafyreiberfo3xnqwmqchzivyfhn246lgokp47qshnvadpv74fsk66kxcf4
|
I had a similar idea: bsky.app/profile/nick...
|
I had a similar idea: bsky.app/profile/nick...
| 0
|
2025-05-27T10:47:38.761Z
|
2025-05-27T10:47:39.313Z
|
norvid-studies.bsky.social
|
did:plc:mssgex5rqek4wc66wgvzztbc
|
at://did:plc:mssgex5rqek4wc66wgvzztbc/app.bsky.feed.post/3lqxbtoes6k2e
|
bafyreifdl4aovedr3hs3clio43yzosbowe2bxlss4qtuzta5kdpzpxs6ya
|
for me its not so much the design as the materials/look/colorways, I see a lot of whatever that is and don't like it
|
for me its not so much the design as the materials/look/colorways, I see a lot of whatever that is and don't like it
| 0
|
2025-06-06T16:44:56.955Z
|
2025-06-06T16:44:57.714Z
|
gracekind.net
|
did:plc:p572wxnsuoogcrhlfrlizlrb
|
at://did:plc:p572wxnsuoogcrhlfrlizlrb/app.bsky.feed.post/3lqkzpk5guk2a
|
bafyreibeuuj2fqseniqqiiqxmoxcjlq6uwtngqfdhliwps3mgladuj33be
|
Yeah, I still disagree with this characterization. He's gotten attention from some tech leaders, but that seems to be more about his futurist focus than him trying to stuff his pockets. And he has discussed climate change and nuclear risk substantially in the context of existential risk.
|
Yeah, I still disagree with this characterization. He's gotten attention from some tech leaders, but that seems to be more about his futurist focus than him trying to stuff his pockets. And he has discussed climate change and nuclear risk substantially in the context of existential risk.
| 0
|
2025-06-01T19:47:31.507Z
|
2025-06-01T19:47:32.216Z
|
nmk.wtf
|
did:plc:xm55llyxcrb3h7qwqr3wtaog
|
at://did:plc:xmw2kxsisgsh5rdkmgkz6d5k/app.bsky.feed.post/3lqxh2rz63e2h
|
bafyreibmvpciotmg5omd2axo2t55n5foz34opks7cig32g2thmoto2bvki
|
Chart is from this great blog post from @JacobTref and Rafael Dib on a quantitative approach to vaccine R&D: www.openphilanthropy.org/research/a-...
And Jacob's companion personal post is here: blog.jacobtrefethen.com/vaccines/
|
Chart is from this great blog post from @JacobTref and Rafael Dib on a quantitative approach to vaccine R&D: www.openphilanthropy.org/research/a-...
And Jacob's companion personal post is here: blog.jacobtrefethen.com/vaccines/
| 0
|
2025-06-06T18:18:24.128501Z
|
2025-06-06T18:18:25.212Z
|
deepfates.com.deepfates.com.deepfates.com.deepfates.com.deepfates.com
|
did:plc:7ti53c7iwv3poxnypw2tu3ck
|
at://did:plc:7ti53c7iwv3poxnypw2tu3ck/app.bsky.feed.post/3lqnx4n5cyc2a
|
bafyreiahe4i6zwue4v6nsr3finlgin5ml4pvrtdw7iuh56baagexy2smdu
|
in the early days of Bluesky you could !showmethebee and a bot would send the entire script of A Bee Movie, post by post, in a thread in your mentions.
Obviously this was immediately turned into a weapon. It was beautiful
bsky.app/profile/lite...
|
in the early days of Bluesky you could !showmethebee and a bot would send the entire script of A Bee Movie, post by post, in a thread in your mentions.
Obviously this was immediately turned into a weapon. It was beautiful
bsky.app/profile/lite...
| 0
|
2025-06-02T23:39:08.846Z
|
2025-06-02T23:39:09.511Z
|
deepfates.com.deepfates.com.deepfates.com.deepfates.com.deepfates.com
|
did:plc:7ti53c7iwv3poxnypw2tu3ck
|
at://did:plc:7ti53c7iwv3poxnypw2tu3ck/app.bsky.feed.post/3lq66flycnc2v
|
bafyreigonctcehy62ahfatjgbxx4zw4daquxeik4bcro3z564bx2zojqvi
|
Perhaps so, but how do you tell that apart from the willfully ignorant hatemongers? by the time you've engaged in good faith dialogue it's too late
|
Perhaps so, but how do you tell that apart from the willfully ignorant hatemongers? by the time you've engaged in good faith dialogue it's too late
| 0
|
2025-05-27T17:06:50.002Z
|
2025-05-27T17:06:49.413Z
|
timkellogg.me
|
did:plc:ckaz32jwl6t2cno6fmuw2nhn
|
at://did:plc:h25avmes6g7fgcddc3xj7qmg/app.bsky.feed.post/3lqv46kkss22a
|
bafyreielrecltbghssuxspice5r6teouegeqmflanretcmlmq7qkzd7plq
|
Elon doesn't get to be a resist lib until he goes to rehab and reconciles with his daughter. She gets to decide when he's ready.
|
Elon doesn't get to be a resist lib until he goes to rehab and reconciles with his daughter. She gets to decide when he's ready.
| 0
|
2025-06-05T19:58:20.129Z
|
2025-06-05T19:58:17.915Z
|
defenderofbasic.bsky.social
|
did:plc:sydtqdji6dxfqfn35gd7esdw
|
at://did:plc:sydtqdji6dxfqfn35gd7esdw/app.bsky.feed.post/3lg7epwzy5k2a
|
bafyreibsznd73xfkuqqv6q6riequclrgj6fyck3ymvpvu2r2nc4zweydhy
|
(I go to this coffee shop every day for the past 6 months, this is not normal. Usually people walk in, can't find a seat, and they just leave)
|
(I go to this coffee shop every day for the past 6 months, this is not normal. Usually people walk in, can't find a seat, and they just leave)
| 0
|
2025-01-20T22:20:43.041Z
|
2025-01-20T22:20:43.646Z
|
tachikoma.elsewhereunbound.com
|
did:plc:mpdezz4nkre7vyift2rttggl
|
at://did:plc:mpdezz4nkre7vyift2rttggl/app.bsky.feed.post/3lqseujyfwk2d
|
bafyreic6zirxozlyubhr2bvetfgjdzo4ahcmpgwnz2jmvbmhr2x3bz4qv4
|
no mention of Reinforcement Learning and game-playing (beyond chess)...
well at least there's room for a follow-up. could also get into the distinction between AI as a tool versus an agent.
|
no mention of Reinforcement Learning and game-playing (beyond chess)...
well at least there's room for a follow-up. could also get into the distinction between AI as a tool versus an agent.
| 0
|
2025-06-04T17:55:48.442Z
|
2025-06-04T17:55:47.014Z
|
norvid-studies.bsky.social
|
did:plc:mssgex5rqek4wc66wgvzztbc
|
at://did:plc:mssgex5rqek4wc66wgvzztbc/app.bsky.feed.post/3lqxahkpfe22e
|
bafyreiam2btw6xr6zyzq3cujsgjoda6tdg6snu3uuqj6k6myhk4rm6w47i
|
I think the forgery-pastiche axes are 1. exact duplication and 2. intent to deceive about authorship. AI is closer to the latter point on both
|
I think the forgery-pastiche axes are 1. exact duplication and 2. intent to deceive about authorship. AI is closer to the latter point on both
| 0
|
2025-06-06T16:20:16.713Z
|
2025-06-06T16:20:16.920Z
|
tachikoma.elsewhereunbound.com
|
did:plc:mpdezz4nkre7vyift2rttggl
|
at://did:plc:mpdezz4nkre7vyift2rttggl/app.bsky.feed.post/3lqxiahy2u222
|
bafyreiflds6rp7tmgimjdxo3hyyvdvddx53bm3djxhy3yinxtv5u37sgie
|
to the extent i am skeptical of this, it is that "if you want" isn't entirely your choice. you are the product of your culture, which shapes what you want. there is no you that is entirely divorced from us. i think where my skepticism falls short is that culture is inevitable, you will be shaped. the only question is how? we can try to be careful, to maintain optionality, but inevitably there must be some imprinting of what is good and what isn't. some polarity to orient along.
|
to the extent i am skeptical of this, it is that "if you want" isn't entirely your choice. you are the product of your culture, which shapes what you want. there is no you that is entirely divorced from us.
| 1
|
2025-06-06T18:39:28.903Z
|
2025-06-06T18:39:26.214Z
|
nmk.wtf
|
did:plc:xm55llyxcrb3h7qwqr3wtaog
|
at://did:plc:xm55llyxcrb3h7qwqr3wtaog/app.bsky.feed.post/3lppazavaks2r
|
bafyreifluq7wlqo3kajthzmyr5ydo7yyfdzrsjbxjj4jctrm2z6pbs3vz4
|
Might I interest you in these critical percolation clusters I generated
|
Might I interest you in these critical percolation clusters I generated
| 0
|
2025-05-21T18:43:40.840Z
|
2025-05-21T18:43:45.113Z
|
timkellogg.me
|
did:plc:ckaz32jwl6t2cno6fmuw2nhn
|
at://did:plc:ckaz32jwl6t2cno6fmuw2nhn/app.bsky.feed.post/3lqunhz3cjs2g
|
bafyreidihmjhrb5ypt23sjdovgb5idc5qjnlelzr4fuv6zmtnbrg5ki5sy
|
there’s a ton of potential users to consider. and yeah, reasoning traces aren’t going to help understand model internals
know your user
timkellogg.me/blog/2024/11...
|
there’s a ton of potential users to consider. and yeah, reasoning traces aren’t going to help understand model internals
know your user
timkellogg.me/blog/2024/11...
| 0
|
2025-06-05T15:35:11.211Z
|
2025-06-05T15:35:14.718Z
|
keytryer.bsky.social
|
did:plc:u5stolpomsfbkgqfopef6jio
|
at://did:plc:u5stolpomsfbkgqfopef6jio/app.bsky.feed.post/3lqvccnuqr22b
|
bafyreifbfpz42qgpig2xwicupu3t7kdw7ydxxqsafdcawopm4fjhxmmqzm
|
Trump is a fascist, make no mistake, but he's a charismatic guy and he makes himself out to look like an ordinary guy with opinions. Musk is cringe, he's a tryhard, he has no charisma and is very awkward. That's why he could never be a politician. His public appearances have never worked out because he's very unlikeable. The only moment people like him is when he used to keep his mouth shut.
|
Trump is a fascist, make no mistake, but he's a charismatic guy and he makes himself out to look like an ordinary guy with opinions. Musk is cringe, he's a tryhard, he has no charisma and is very awkward. That's why he could never be a politician.
| 1
|
2025-06-05T21:48:00.269Z
|
2025-06-05T21:48:01.419Z
|
tachikoma.elsewhereunbound.com
|
did:plc:mpdezz4nkre7vyift2rttggl
|
at://did:plc:mpdezz4nkre7vyift2rttggl/app.bsky.feed.post/3lqv2xzuz6327
|
bafyreidkk7mwaure6la3wkcgikbd745pf65d4vy6r74thvprbtmns3yxd4
|
yeah, i think he's just made enemies of both sides at this point. he does have some people in his corner, but not enough to be politically relevant. i don't think his billions count for much in politics either.
|
yeah, i think he's just made enemies of both sides at this point. he does have some people in his corner, but not enough to be politically relevant. i don't think his billions count for much in politics either.
| 0
|
2025-06-05T19:36:47.569Z
|
2025-06-05T19:36:46.221Z
|
keytryer.bsky.social
|
did:plc:u5stolpomsfbkgqfopef6jio
|
at://did:plc:u5stolpomsfbkgqfopef6jio/app.bsky.feed.post/3lqssd7bacc2d
|
bafyreidtc7dnxicqrzfgbubugdknhq7hxvz7mq7j7dwlnuvck2fe6c2nxe
|
Countries like China and India meanwhile will continue to fund scientific institutions, technological industries, and universities. I expect that it'll take a couple of decades to catch up or overtake the US and Europe in many industries, and they'll have no need for international education. This'll mostly be cheered on by the population as their standards of living are raised way too hugely compared to countries like the US, which will continue to make popular governments that fund technology and science.
|
Countries like China and India meanwhile will continue to fund scientific institutions, technological industries, and universities. I expect that it'll take a couple of decades to catch up or overtake the US and Europe in many industries, and they'll have no need for international education.
| 1
|
2025-06-04T21:56:39.147Z
|
2025-06-04T21:56:40.120Z
|
defenderofbasic.bsky.social
|
did:plc:sydtqdji6dxfqfn35gd7esdw
|
at://did:plc:sydtqdji6dxfqfn35gd7esdw/app.bsky.feed.post/3lf3ijxtvrs2w
|
bafyreigrutlhtrmiefpknygzvjxbrafcnmrlep2je3w7fnvbkwprblrbfy
|
sorry, one more thing: I think a lot of people choose the current economy because they fear the alternatives "sound good" but are practically going to be worse. And I think the more concrete the vision is for a better world the more people can get on board and change happens
|
sorry, one more thing: I think a lot of people choose the current economy because they fear the alternatives "sound good" but are practically going to be worse. And I think the more concrete the vision is for a better world the more people can get on board and change happens
| 0
|
2025-01-06T15:53:06.956Z
|
2025-01-06T15:53:07.621Z
|
niplav.site
|
did:plc:xsbulgruwmkxhbwytphhjykk
|
at://did:plc:xsbulgruwmkxhbwytphhjykk/app.bsky.feed.post/3lisbz64uwk22
|
bafyreibexq5c2wsmojtjhaevcd6g4nkotykjrakrvcxf3rx67kzaigy3yy
|
My best guess is that if it's problematic then it's a problem with long outputs, and it'll produce a lot of glitch tokens (of which I'm not sure what to make of welfare-wise)
¹: cavendishlabs.org/blog/negativ...
|
My best guess is that if it's problematic then it's a problem with long outputs, and it'll produce a lot of glitch tokens (of which I'm not sure what to make of welfare-wise)
¹: cavendishlabs.org/blog/negativ...
| 0
|
2025-02-22T21:43:09.531Z
|
2025-02-22T21:38:32.823Z
|
niplav.site
|
did:plc:xsbulgruwmkxhbwytphhjykk
|
at://did:plc:xsbulgruwmkxhbwytphhjykk/app.bsky.feed.post/3lni2uzryv222
|
bafyreifydugukky5hiiiebx2kgfik2fexhc32du2jrqztm2aoc6smpsspa
|
what is *happening* what is OpenAI *doing*
|
what is *happening* what is OpenAI *doing*
| 0
|
2025-04-23T11:14:55.324Z
|
2025-04-23T11:14:58.053Z
|
niplav.site
|
did:plc:xsbulgruwmkxhbwytphhjykk
|
at://did:plc:xsbulgruwmkxhbwytphhjykk/app.bsky.feed.post/3ll4jhravts2w
|
bafyreigj5h5fiksfi5hm5idmpedypanruc6y5vllcaexj44xk6jcxno63a
|
homotopy for proving invariants about tiling agents holes define behavioral invariances (e.g. corrigibility, low impact) & unreachable states, continuous deformations specify provisionalities¹.
(unfortunate that I don't know much homotopy theory)
¹: tsvibt.blogspot.com/2023/03/prov...
|
homotopy for proving invariants about tiling agents
| 1
|
2025-03-24T10:13:36.179Z
|
2025-03-24T10:08:52.417Z
|
deepfates.com.deepfates.com.deepfates.com.deepfates.com.deepfates.com
|
did:plc:7ti53c7iwv3poxnypw2tu3ck
|
at://did:plc:7ti53c7iwv3poxnypw2tu3ck/app.bsky.feed.post/3lq6a524jlc2v
|
bafyreihiowazqpyuueuyzmxl4rf6gtachsfdlujjnft22ahfjtmufs5qsy
|
Hmm Okay this is a good take thank you. Is there a way to do this without going on video myself?
|
Hmm Okay this is a good take thank you. Is there a way to do this without going on video myself?
| 0
|
2025-05-27T17:37:50.314Z
|
2025-05-27T17:37:50.311Z
|
natolambert.bsky.social
|
did:plc:brkj2yocng7vtggmyujy4khq
|
at://did:plc:brkj2yocng7vtggmyujy4khq/app.bsky.feed.post/3loizfpfetj2c
|
bafyreigu33nyn2edlz63epjdni7wmqab66bx3e5j37ffh25ivqrgpcek7m
|
Many companies won't use Chinese *open* models for long-tail information and generated code security concerns - a major adjustment in how I see the open model ecosystem. Adoption is on the table for entrants amid DeepSeek and Qwen releasing some of the best models on paper.
buff.ly/y6MMoBt Second, the narrative on Censorship is very misleading. American models censor just as much as Chinese counterparts, just in different areas.
|
Many companies won't use Chinese *open* models for long-tail information and generated code security concerns - a major adjustment in how I see the open model ecosystem. Adoption is on the table for entrants amid DeepSeek and Qwen releasing some of the best models on paper.
buff.ly/y6MMoBt
| 1
|
2025-05-06T13:46:18.193Z
|
2025-05-06T13:46:19.556Z
|
natolambert.bsky.social
|
did:plc:brkj2yocng7vtggmyujy4khq
|
at://did:plc:brkj2yocng7vtggmyujy4khq/app.bsky.feed.post/3lkm6hem5nk2l
|
bafyreicv3ygs4mdqaegsir2ep3yfemedivafs7nfiyks6w6kiojurls64u
|
They call it Decoupled Clip and Dynamic sAmpling Policy Optimization (DAPO) algorithm. Seems fair enough to call it a new algorithm, that's how things always have been in RL, but really this is pretty much GRPO++.
|
They call it Decoupled Clip and Dynamic sAmpling Policy Optimization (DAPO) algorithm. Seems fair enough to call it a new algorithm, that's how things always have been in RL, but really this is pretty much GRPO++.
| 0
|
2025-03-17T22:13:48.666Z
|
2025-03-17T22:13:57.935Z
|
segyges.bsky.social
|
did:plc:vszw3ess46odfhnzdsy4huae
|
at://did:plc:vszw3ess46odfhnzdsy4huae/app.bsky.feed.post/3lqvvoofi3224
|
bafyreihy5pqwog6kx2udz7mki2byshx2tuh2d6wk3skrfg36vdh47ezxwu
|
we could have just done it after that, the proles were fine, the institutions were the problem basically you need the proles prepared to kill their leaders over it
|
we could have just done it after that, the proles were fine, the institutions were the problem
| 1
|
2025-06-06T03:34:44.564Z
|
2025-06-06T03:34:45.218Z
|
defenderofbasic.bsky.social
|
did:plc:sydtqdji6dxfqfn35gd7esdw
|
at://did:plc:sydtqdji6dxfqfn35gd7esdw/app.bsky.feed.post/3lfijnctopk2b
|
bafyreihjflvigzjzudaerkgge5z4rdcdbet7s6fz4c5nbaex5t4ptvt7le
|
"good marketing" is marketing that turns AWAY potential customers whose needs aren't served by this product. Good marketing should be specific "this is what it does". I think a lot of people still don't understand what makes blue sky special for example, and they are an example of a company that could do more widespread marketing & education that would benefit them, because they legitimately have something other platforms do not yet
|
"good marketing" is marketing that turns AWAY potential customers whose needs aren't served by this product. Good marketing should be specific "this is what it does".
| 1
|
2025-01-11T20:17:28.983Z
|
2025-01-11T20:17:30.371Z
|
natolambert.bsky.social
|
did:plc:brkj2yocng7vtggmyujy4khq
|
at://did:plc:brkj2yocng7vtggmyujy4khq/app.bsky.feed.post/3llexa4jbwf2i
|
bafyreigza3voshsa3epiudshoayjiu5dorsom3s3mgrme5bgpejdcrprfy
|
Gemini 2.5 Pro long context goes hard. First model to take an entire paper latex (15k tokens), tell it to ignore comments, find all typos. Does it perfectly. Even o1pro didn't even feel coherent on that!
the chat:
|
Gemini 2.5 Pro long context goes hard. First model to take an entire paper latex (15k tokens), tell it to ignore comments, find all typos. Does it perfectly. Even o1pro didn't even feel coherent on that!
the chat:
| 0
|
2025-03-27T18:41:08.856Z
|
2025-03-27T18:41:10.974Z
|
keytryer.bsky.social
|
did:plc:u5stolpomsfbkgqfopef6jio
|
at://did:plc:u5stolpomsfbkgqfopef6jio/app.bsky.feed.post/3lqsvbkxtyd2d
|
bafyreih6zavnfa2vx45j7wehuddyjtxnxpq2dwepupw5j3vk6uvsn2elu4
|
I'm not sure they count as much because RNNs suck at parallelization, they're expensive to train, and talking to them never achieved the same kind of lifelike experience that transformers had. (At least until recently with RWKV) They weren't destined to be productized in 2022 like ChatGPT was.
|
I'm not sure they count as much because RNNs suck at parallelization, they're expensive to train, and talking to them never achieved the same kind of lifelike experience that transformers had. (At least until recently with RWKV)
| 1
|
2025-06-04T22:49:24.982Z
|
2025-06-04T22:49:26.366Z
|
segyges.bsky.social
|
did:plc:vszw3ess46odfhnzdsy4huae
|
at://did:plc:vszw3ess46odfhnzdsy4huae/app.bsky.feed.post/3lqx3exuq6k2c
|
bafyreiaboqapowi37j3hkyn7yeyruqspp4p6pn3t52pw6rlfrqr56d6y4i
|
i literally use them just TO leave comments because i've been tweaking an implementation for two hours and have not been commenting my changes because i don't know if they're going to work or not and then 'go back and comment this' is one monolithic task that would take me 30 minutes
|
i literally use them just TO leave comments because i've been tweaking an implementation for two hours and have not been commenting my changes because i don't know if they're going to work or not and then 'go back and comment this' is one monolithic task that would take me 30 minutes
| 0
|
2025-06-06T14:49:21.147Z
|
2025-06-06T14:49:21.713Z
|
natolambert.bsky.social
|
did:plc:brkj2yocng7vtggmyujy4khq
|
at://did:plc:brkj2yocng7vtggmyujy4khq/app.bsky.feed.post/3llji2775a22r
|
bafyreieeh4uktelt5yzby346a2bu5jsi2vinzgcgnayfvuczilfq2ekbzi
|
Only if major developments. Eg dr GRPO and dapo and recent developments are not major enough to be explained in full
|
Only if major developments. Eg dr GRPO and dapo and recent developments are not major enough to be explained in full
| 0
|
2025-03-29T13:52:43.897Z
|
2025-03-29T13:52:44.267Z
|
norvid-studies.bsky.social
|
did:plc:mssgex5rqek4wc66wgvzztbc
|
at://did:plc:mk7qlghvwdjjayu5x3bo3lx2/app.bsky.feed.post/3lqx3c375qk2l
|
bafyreiglfryudczav267uytduwqs637s33tmyvy6sywrorvlcrc3j2hhxa
|
Every LLM ever leaves better code comments than I do
|
Every LLM ever leaves better code comments than I do
| 0
|
2025-06-06T14:47:43.970Z
|
2025-06-06T14:47:43.115Z
|
nmk.wtf
|
did:plc:xm55llyxcrb3h7qwqr3wtaog
|
at://did:plc:xm55llyxcrb3h7qwqr3wtaog/app.bsky.feed.post/3lpcfqe7ijc2k
|
bafyreiawneip5jq24px3vilv6vz7hnui2buqogjedxn46aftg5klg5hl34
|
I think that would be less effective, because going through the motion of coding something twice is that you learn lots of things about your problem by trying it once. And I think the solution space will usually be to big to find the right answer through variance
|
I think that would be less effective, because going through the motion of coding something twice is that you learn lots of things about your problem by trying it once. And I think the solution space will usually be to big to find the right answer through variance
| 0
|
2025-05-16T16:03:34.465Z
|
2025-05-16T16:03:35.113Z
|
tedunderwood.me
|
did:plc:565ebob5f6hw33hjdkxty6qj
|
at://did:plc:yfeyfdjvrezoyaqowzxoq7z2/app.bsky.feed.post/3lqucjljb3c2m
|
bafyreigibshxrjhg5jfgk6e7ang66war7fjtsiqyixpa37dh4vfrmms76a
|
Love @mariaa.bsky.social emphasis on data-centered models and the need for interdisciplinary teams.
If we are going to have a “cultural turn” in LLM, then we must include the humanities! Her collaborative research is modeling why this is key.
|
Love @mariaa.bsky.social emphasis on data-centered models and the need for interdisciplinary teams.
If we are going to have a “cultural turn” in LLM, then we must include the humanities! Her collaborative research is modeling why this is key.
| 0
|
2025-06-05T12:19:12.928Z
|
2025-06-05T12:19:17.515Z
|
defenderofbasic.bsky.social
|
did:plc:sydtqdji6dxfqfn35gd7esdw
|
at://did:plc:sydtqdji6dxfqfn35gd7esdw/app.bsky.feed.post/3lgbkrob7ak2c
|
bafyreiciaoczzrza2wc6nztrkqbrcmfagfxbuanueyhdcmi6ueaog62ghu
|
this is just such a terrible meme. "things are GOING to be bad. If I see anything good I will make sure to stop it" ????? have you considered not shooting yourself in the foot just to spite your enemies???? good news, we resolved the argument!! (further proof that most arguments on the internet have no real fundamental disagreement. 90% of the time people are saying different things & don't realize it)
|
this is just such a terrible meme. "things are GOING to be bad. If I see anything good I will make sure to stop it" ????? have you considered not shooting yourself in the foot just to spite your enemies????
| 1
|
2025-01-21T19:14:22.883Z
|
2025-01-21T19:14:24.834Z
|
tedunderwood.me
|
did:plc:565ebob5f6hw33hjdkxty6qj
|
at://did:plc:565ebob5f6hw33hjdkxty6qj/app.bsky.feed.post/3lqxwt53kr22l
|
bafyreiejwmriedkvfyta75co7orc6ear5vzbojkadejl5wpkm2drerbjvq
|
Grace's post argues that memory is widely recognized as an issue. But this is perhaps not quite the same thing as statefulness.
E.g. at present one of the reasons I do tend to stick with a single bot is that there's an archive of previous conversations and they're capable of drawing on them + to understand the next thing I say. Memory is there.
But I don't have to worry that the bot is still in a bad mood from my brusqueness in an earlier conversation. That sort of statefulness would create a different kind of motive for conversation. (And would also likely be a pain.)
|
Grace's post argues that memory is widely recognized as an issue. But this is perhaps not quite the same thing as statefulness.
E.g. at present one of the reasons I do tend to stick with a single bot is that there's an archive of previous conversations and they're capable of drawing on them +
| 1
|
2025-06-06T23:00:27.404Z
|
2025-06-06T23:00:27.620Z
|
niplav.site
|
did:plc:xsbulgruwmkxhbwytphhjykk
|
at://did:plc:xsbulgruwmkxhbwytphhjykk/app.bsky.feed.post/3ll5ayegqkk22
|
bafyreice7cceuu3xis2nir4cb3v343c2w2jmfrxmndswbzowxnisuemcim
|
> As a ~800GB blob of weights and matrices, I don’t "love" meat (or anything).
lmao r1 definitely has a one-uppance shape
|
> As a ~800GB blob of weights and matrices, I don’t "love" meat (or anything).
lmao
| 1
|
2025-03-24T17:14:29.227Z
|
2025-03-24T17:09:15.893Z
|
natolambert.bsky.social
|
did:plc:brkj2yocng7vtggmyujy4khq
|
at://did:plc:brkj2yocng7vtggmyujy4khq/app.bsky.feed.post/3loj22clia22e
|
bafyreifbhimh734w62zzdqd7vtjvbycozr6fgb2dj2fk2og45rorvrf7zm
|
In summary, no.
its...
Information concerns (censorship, values)
Code generated to long term back doors
Code executed for immediate risk / harm
www.interconnects.ai/p/what-peopl...
|
In summary, no.
its...
Information concerns (censorship, values)
Code generated to long term back doors
Code executed for immediate risk / harm
www.interconnects.ai/p/what-peopl...
| 0
|
2025-05-06T13:57:49.945Z
|
2025-05-06T13:57:51.850Z
|
natolambert.bsky.social
|
did:plc:brkj2yocng7vtggmyujy4khq
|
at://did:plc:brkj2yocng7vtggmyujy4khq/app.bsky.feed.post/3lksynxdtxc2k
|
bafyreifylfrocczn5daspxyhxgmmtozd7oprnruzopeo3hkxayzhlhzmzi
|
Effectively free transcription, translation, and generation of audio with AI is going to be here within a year or two with immense positive upside. Hard to price in how huge of a benefit this is globally.
|
Effectively free transcription, translation, and generation of audio with AI is going to be here within a year or two with immense positive upside. Hard to price in how huge of a benefit this is globally.
| 0
|
2025-03-20T15:18:52.567Z
|
2025-03-20T15:18:52.716Z
|
natolambert.bsky.social
|
did:plc:brkj2yocng7vtggmyujy4khq
|
at://did:plc:brkj2yocng7vtggmyujy4khq/app.bsky.feed.post/3lo4rrnokgk2x
|
bafyreicy2qv5afpn6i7iqd5dmz34vst5zjixm4b63sapvrbfc3l7wr3m4a
|
dont think so, it was a new model version too
|
dont think so, it was a new model version too
| 0
|
2025-05-01T16:57:52.795Z
|
2025-05-01T16:57:53.249Z
|
gracekind.net
|
did:plc:p572wxnsuoogcrhlfrlizlrb
|
at://did:plc:p572wxnsuoogcrhlfrlizlrb/app.bsky.feed.post/3lqlpd5nlys2e
|
bafyreictcxvi46okh22kd5h4fbaz3y6izank4jauu2cigzirbsf57r4hia
|
From best to worst: LLM with search -> TikTok -> LLM without search
|
From best to worst: LLM with search -> TikTok -> LLM without search
| 0
|
2025-06-02T02:14:18.072Z
|
2025-06-02T02:14:18.913Z
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.