On Wednesday, October 07, 2020, at 10:57 AM, James Bowery wrote: > How many times have I stated in this group that big language modelers > bragging about their parameters counts worse than cottoncandyfluffpuffery? > Reducing the parameter counts by a factor of 1000 and achieving comparable > performance on just about _any_ plausible language benchmark is damning of > such hypercottoncandyfluffpuffery.
Because it doesn't matter much if we make GPT-3 1,000 times a smaller model!!!! It will still be a stupid GPT. It won't be SMARTER. It'll only take up less memory on your storage lol and maybe run faster ex. 10x faster. Still stupid, no improvement!!! Do note GPT is not a bad model, it is certainly closer to AGI - I don't understand why Ben calls it non-AGI because ALL AI find patterns (non-hand-crafted patterns) on their own, except for the GOFAI types. GPT seems to be one of the top here, can you give me a better text predictor? I think not. Maybe you have one but it is too slow or big of a model, well, let me know it hasn't digested enough data in time/mem and must be optimized then. On Wednesday, October 07, 2020, at 11:12 AM, James Bowery wrote: > And, ID, in the final analysis when you say "if it scores better it scores > better" you are begging the question. > > That's less than not interesting. It's a distraction. > > Here's another way to talk about the situation that isn't a distraction: > > "If it predicts better, it predicts better." > > And now think about what it means to "predict better". > > What is the essential function of a lossless compressor? To make its thoughts become real, to turn Earth into patterns so it lives longer, by making general purpose cube-shaped mobile devices like phones and trucks, tables, Kleenex boxes, beds, etc. These patterns that are repeating in clones or life length cause Earth to live longer / repeat. Patterns make patterns emerge, more you have the faster. 3rd evaluation is compression in mind, 2nd is in world, 1st evaluation is pattern size of ex. Earth i.e. size/ how long it maintains form, it is ranking the pattern/immortality. All else means nothing if it dies after 0.0001 nanoseconds. It's ALL about immortality / clones / etc ///// patterns! On Friday, October 09, 2020, at 2:34 PM, James Bowery wrote: > I can't help but suspect that the _real_ reason we're 10 years into Legg's > Measures of Machine Intelligence > <https://www.youtube.com/watch?v=0ghzG14dT-w> and still not using it, despite > Legg co-founding Google DeepMind, is the Boaspawn infestation of Google, > secondary to network effect rents attracting all manner of rentiers highly > evolved to sniff out economic rent from 10,000 miles away -- sending them > into a frenzy. Interesting, so that's why the new hutter prize? 10x the compute every 10 years so 10x the intelligence we should have. ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T6761a13445e5864b-M8b45f45c273e24805a3a1387 Delivery options: https://agi.topicbox.com/groups/agi/subscription
