I think it's an absurd question in some sense
LLMs perform maximization of conditional probability of the next word being correct.
Suppose they get to the point where they do that with 100% accuracy. How can you tell the difference between that and "Reasoning"? You can't. So then the question of whether they are "Reasoning" or not is religious, not quantitative.
Really? Are you sure ?
His article can basically be summed up as don’t believe AI hype from Sama. It’s not particularly well written, he’s no Nabokov. ChatGPT bangs out stuff like this effortlessly.
Here, I did it for you
https://chatgpt.com/share/67019a6c-453c-8006-88aa-6f32435492...
where are you getting that information that it's not the meaning of the word that OpenAI talked about originally? Did they ever promise that the model weights, or exact architecture would be open? I would like to see that promise. What exactly in your opnion did they promise that they have violated?
OpenAI and Anthropic for sure have products and that's great.
However, these products are pretty far from a super intelligence.
The bet SSI is making is that by not focusing on products and focusing directly on building a super intelligence, they can leapfrog all these other firms
Now, if you assign any reasonable non zero probability to them succeeding,the expected value of this investment is infinity. It's definitely a very risky investment, but risk and expected value are two different things.
Sure. Expected value and risk are different things. Clearly, such an investment is very risky. It’s easy to imagine SSI failing. But if you allow even a 1% chance of success here, the expected value is infinite
you are missing the point. SSI believes that it can build a super intelligence. Regardless of whether you personally buy into that or not, the expected value of such an investment is infinity effectively. 5 billion dollar valuation is a steal
Some CEOs are bad at running companies and some founders are also bad. Boeing example is weak for that reason. I just don't think this statement is empirically true at all "if you’re smart and intrinsically curious, these deep dives aren’t really that much time and effort"
I disagree.
Some things are just complex and inherently outside of the CEOs competency.
They have to be able to build a management team of properly incentivized highly skilled professionals. Time in the day is limited, if you are delving into details on X and Y, that means at that same time you are not spending time on Z.
Come on. There is no way CEO of a huge company like Amazon can really know every detail. At some point he’s gonna just have to trust someone at the company.
all this implies trust in the management team, right? to tell Bezos that it's really solved. The opposite of what the essay is arguing, that these managers tend to be fakers and can't be trusted.
well, then founder mode is a meaningless thing because the essay seemed to suggest doing the opposite of "think he just forwarded the email to one of the S-team and it would get passed down the org chart until it reached the relevant manager"
You cannot get involved in every decision as a CEO of a large company.
Suppose you did a skip level meeting. Who is going to gaslight you more, the lower level person or your direct report with a massive stock award who is incentivized for the company to succeed? far from clear.
correct. the last thing that a company needs is a CEO who doesn't empower his direct reports, can't hire good ones, thinks they are all fakers and tries to meddle in every detail, instead of thinking about where the company is going and where it should go and how it should get there.
Agree, and I feel like when people say “sleeping” it sounds innocuous. But what do these people do if they need to go number 2 in the middle of the night? There is a reason we have campsites, sanitation and just basic limits on insane behavior. “Where should they go” is something I hear often. Well, guess what, if we create significant disincentives to just doing whatever you feel like just because, at least some fraction of these people will find somewhere to go, a family member, a shelter, somewhere. And we can build more shelters, sure, but this has gotten out of control.
The way you are phrasing your disagreement with prevailing consensus is just semantics. Ok, let’s say animals are not 0% conscious, let’s say a certain animal is 24% conscious. It’s reasonable to say that, in comparison to a human, this animal is not conscious. Clearly, my cat exhibits some awareness of what’s going on, it’s not totally mindless. But to call it conscious in the way that we apply this term to humans is a stretch.