text
stringlengths
0
2.35k
[2977.68 --> 2978.64] my unpopular opinion
[2978.64 --> 2979.36] if you want to,
[2979.46 --> 2980.60] but it's kind of my opinion
[2980.60 --> 2981.60] that if Ghidra
[2981.60 --> 2982.48] had never appeared,
[2983.00 --> 2984.08] then IDAPro
[2984.08 --> 2984.72] would basically
[2984.72 --> 2985.24] have stayed
[2985.24 --> 2986.62] kind of the same
[2986.62 --> 2987.64] like for the next decade
[2987.64 --> 2988.10] or so
[2988.10 --> 2989.34] because the developers
[2989.34 --> 2990.50] had kind of no incentive
[2990.50 --> 2991.80] to make it
[2991.80 --> 2992.70] significantly better
[2992.70 --> 2993.38] because just
[2993.38 --> 2994.60] they had no competition there.
[2995.08 --> 2996.40] The way that the compiler
[2996.40 --> 2997.58] is working at the moment
[2997.58 --> 2999.02] as far as I understand
[2999.02 --> 2999.88] is purely
[2999.88 --> 3001.34] through algorithmic means
[3001.34 --> 3002.56] and they do not use
[3002.56 --> 3003.58] any form of
[3003.58 --> 3005.16] machine learning.
[3005.48 --> 3006.14] There is no AI
[3006.14 --> 3007.08] applied to
[3007.08 --> 3007.72] their
[3007.72 --> 3009.42] decompilation process.
[3010.22 --> 3011.18] Maybe they have
[3011.18 --> 3012.30] started working on this
[3012.30 --> 3012.64] but
[3012.64 --> 3014.64] as a company
[3014.64 --> 3015.32] and maybe I'm
[3015.32 --> 3015.90] totally mistaken
[3015.90 --> 3016.32] about this,
[3016.38 --> 3017.06] I don't work there
[3017.06 --> 3018.02] but Hexray
[3018.02 --> 3018.74] doesn't strike me
[3018.74 --> 3019.34] as a company
[3019.34 --> 3020.08] that would be doing
[3020.08 --> 3021.26] groundbreaking R&D
[3021.26 --> 3022.24] that would
[3022.24 --> 3022.86] maybe
[3022.86 --> 3024.30] prepare for the
[3024.30 --> 3025.18] the next generation
[3025.18 --> 3026.08] of decompilers.
[3026.18 --> 3026.66] I think they would
[3026.66 --> 3027.90] rather make incremental
[3027.90 --> 3029.12] changes on their
[3029.12 --> 3029.78] existing product
[3029.78 --> 3030.54] to make it slightly
[3030.54 --> 3031.54] better year after year.
[3032.24 --> 3033.24] I think Ghidra
[3033.24 --> 3034.40] which has a
[3034.40 --> 3035.32] decompiler as well
[3035.32 --> 3036.50] is open source
[3036.50 --> 3037.38] and I think this
[3037.38 --> 3038.24] decompiler is doing
[3038.24 --> 3039.52] pretty well too
[3039.52 --> 3040.84] but I don't think
[3040.84 --> 3041.64] it uses AI
[3041.64 --> 3042.98] in any shape
[3042.98 --> 3043.70] or form either
[3043.70 --> 3045.42] and I'm not aware
[3045.42 --> 3046.52] of any plans
[3046.52 --> 3047.76] to like to start
[3047.76 --> 3048.44] working on this.
[3048.92 --> 3049.48] Probably
[3049.48 --> 3050.98] developing some
[3050.98 --> 3051.84] AI product
[3051.84 --> 3052.96] focused on reverse
[3052.96 --> 3053.34] engineering
[3053.34 --> 3053.96] we require
[3053.96 --> 3054.70] some very
[3054.70 --> 3056.22] specific AI
[3056.22 --> 3056.74] knowledge
[3056.74 --> 3058.90] and my feeling
[3058.90 --> 3059.60] is I don't know
[3059.60 --> 3060.54] everyone in the
[3060.54 --> 3061.08] security field
[3061.08 --> 3062.46] but I tend to
[3062.46 --> 3063.70] like my perception
[3063.70 --> 3064.44] of this is that
[3064.44 --> 3065.14] we have people
[3065.14 --> 3065.68] that are extremely
[3065.68 --> 3066.32] focused and
[3066.32 --> 3067.10] extremely talented
[3067.10 --> 3067.60] and skilled
[3067.60 --> 3069.00] in the specific
[3069.00 --> 3070.02] field of
[3070.02 --> 3070.54] cybersecurity
[3070.54 --> 3071.08] that they are
[3071.08 --> 3071.62] working on
[3071.62 --> 3072.98] but I cannot
[3072.98 --> 3074.30] recall anyone
[3074.30 --> 3074.96] that I've met
[3074.96 --> 3076.22] that was both
[3076.22 --> 3077.26] a great researcher
[3077.26 --> 3077.92] or a great