I can see a high value startup, that will provide Human Intelligence with real Humans, locked in the room, with no network, books, LLMs and monitored 24x7 with cameras.
Its less of open debate would say, and although superposition [1] is interesting, as a way to explain power of some effects, it is clear they are right now closer to Stochastic Parrots than AGI.
Why do I say that? Because you can trivially beat most guardrails, simply by encoding your prompt in base64 for example. :-) Just word matching...no real understanding.
>> When the inevitable happens I really do think it going to be pretty bad this time
Since all these AI companies, are offering a stake to the US Government so their insane, unrecoverable Capex can get its bailout, I can only imagine this will be pretty bad...for the US Taxpayer...
Data Centers in Space are a practical engineering impossibility, as well as making no economic sense. Engineering and the laws of physics get on the way.
Just because Scott Manley refuses to call that out, so he can do another eight videos about it, don´t stop listening to somebody with the feet on the ground:
The best benchmarks are the ones you create yourself.
Its not my experience Opus is leagues ahead or even superior, but in any case, since GPT 5.5 has Instant, Medium, High, Extra High and Pro...Should the comparison be with GPT on Pro, instead of Extra High as it seems to be the case in the table?
"...In this paper, we propose Autoregressive Boltzmann Generators (ArBG) -- a novel autoregressive modelling framework -- that overcomes these limitations by departing from the flow-based BG paradigm. ArBG circumvents the topological constraints of flows and enables sequential inference-time interventions, while offering enhanced scalability by leveraging architectures effective in Large Language Models...
...We empirically demonstrate that ArBG leads to significant improvements over flow-based models across all benchmarks, but particularly in larger peptide systems such as the 10-residue Chignolin...
...Furthermore, we introduce Robin, a 132 million parameter transferable model trained with the ArBG framework which improves over the previous state-of-the-art, reducing the zero-shot energy error, E-W, on 8-residue systems by over 60. The code can be found at the following link: " - https://github.com/danyalrehman/autobg