@MarioNawfal He is absolutely correct. The destruction AI brought to this world will show up in about 10 years when current generation matures. Our brains need constant training to reason even at much later ages like over 70 !
@GaryMarcus LLMs have zero reasoning capabilities. They are just next token predictors which is pretty dumb when you think how human brain really works. There is no gradient decent in human brains :)
@MrEwanMorrison Addiction to AI can be much worse than the drug addiction for school age children. At young age, brains properly develop only with constant self involvement in reasoning. This crisis will show up very soon everywhere in the world.
@GaryMarcus@geoffreyhinton It is correct. What is important is these tech companies think they can use some stupid AI agents to fix it. If LLMs regurgitate the data how can it even figure out how to screen it?
Actually this is very intuitive and absolutely no surprise here. LLMs can retain more information and spit out using statistics (not intelligence) that will be above average because averages will have lot of less educated people. LLMs statistical regurgitation of information can never exceed any highly intelligent person.
@MrEwanMorrison Not only that, no company has proven yet models get improved significantly with spending more on electricity to train. They are wasting money on products that give only marginal improvements.
@damianplayer Total BS. For an LLM, the weights in the model is everything ! If someone has a model with trained weights, every single bit of information can be extracted from it.
@GaryMarcus This has to be totally true. There is no material available in this world for LLMs to use for do the training for larger context model. Whoever doesn't steal everything available will lose whatever the edge they want against other models !
@knowRowan Proof is right here. Just watch the comments of all tech CEOs and see how delusional they have become. They still can't admit AI doesn't change the world.
@cosmicfibretion This nonsense getting to a such ridiculous level Congress must act. How bad is when AI is hallucinate during a diagnosis? AI must be absolutely banned from science, medicine and psychotherapy. AI can do certain limited things in other areas.
@heynavtoor Of course all of the tech companies do it for very simple reason, there is not enough non copy-righted material left to do training with a bigger context length! Not stealing is the not possible thing !
@kakashiii111 All of them will pay the price. AI has nothing to offer except writing poems and creating funny videos and some marginal cases for writing software with so many errors. This is absolute disgrace to any intelligence to pump this stupid narrative.
@DividendTalks You didn't miss anything. They are wasting capital like there is no tomorrow. It all will show up in future earnings because capital expenses are on a depreciating schedule.
@oguzerkan MSFT valuation must be much lower from here, they have no credible plan for anything new except to have legacy businesses going forward. So, capital they waste now is not yet reflected in the stock price.