I understood the video just fine. You don't understand the difference between performance and capability. You are also confusing training time with quality of training. AI models (models,not LLMs) continue to improve in *capability*, and nobody has suggested they have stopped improving or found any reason to believe that this exponential upward trend in capability has suddenly plateud.

So the "hype" was real (news flash ... if it was real it wasn't hype) but now ... Just a few months ago ... this year ... much more capable models didn't get released? That's an absurd claim. It sounds like you simply didn't know what the word capable means, and think it is about improvement in a single domain. Capability is about models used across domains. If a model doesn't improve at all in a single application domain (say code generation) the capability still improves dramatically if it improves in multiple other domains. You are also still making the same mistake they are making by equating the LLM with the model. They are talking about scaling, which has to do with model training time ... not capability ... then concluding that because scaling doesn't reduce training time exponentially that scaling doesn't improve model capability exponentially. Again, nobody ever claimed that it does.

My university did the opposite, and to this day I think it was good.

At the end of the first semester, you were given a test in the three most important subjects for your field. If you failed, you could re-take it after the summer break, before next semester starts. If you failed it again, you could repeat the first semester and try a third and final time. Fail that, you're out.

It was hard, too. In my class, we were just over a hundred when we started. We were 28 in the 2nd semester. But of those 28, almost everyone made it to the end. Everyone who would've failed somewhere along the way had already been weeded out.

It was brutal, it was stressful - but IMHO it was the best way to do it and be respectful of both student time and university resources.

We had that in 1st grade. Not university, school. So students starting university have now regressed to the level of 6-year-olds ?

I get that it's stressful. I don't get how "managing stress" has vanished from the requirements of getting a degree.

Investors are chasing a company that achieves some kind of AI singularity. Let's set aside the fact that there's no reason to believe there is anything but diminishing marginal returns by making marginal refinements to current frontier models. Let's imagine someone hits the jackpot and gets, not even AGI, but a system that's as far ahead of today's frontier model are ahead of 2020's GPT 2.0 in performance.

Globally AI revenues are 150 billion, against a cumulative burn rate of 450 billion. A model that is a generation ahead of others would almost certainly capture the lion's share of that revenue.

If AGI magically appears as a Sam Altman has promised investors it will, a hundred million is way too low. Add, maybe, another zero to the revenues.

Conservatively, a safer assumption is that frontier models will get marginally better based on refinements in training and reinforcement and the other bits and bobs that go into these systems. The nobody is winning the lion's share of anything, at least overnight. But you have to define "safe". By "safe" I mean unlikely to lose money. But some investors are clearly defining "safe" as "having the greatest chance of owning a piece of the biggest thing ever."

I'm not following this super-closely, but if Anthropic is pursuing adding multi-step model based reasoning to their system, that could be the basis of a generational leap in capability. But if that is an approach that looks like it has a chance of working, then their competitors are no doubt pursuing the same thing. In that case you'd expect the revenue pie to grow as the scope of model utilty increases, but that growth to be split among several competitors. This could credibly result in a revenue stream for some of them that is as big as the entire industry's revenue stream today. But there's going to be hell to pay on the data center impacts end of things.

First of all you don't know what they are doing. Secondly, you failed to grasp the major failure in your reasoning again. If the way they do it is inaccurate that in no way accounts for any aberration in the data.

So your theory is that the OS fingerprinting became significantly more unreliable for a short period of time, and then returned to the previous degree of unreliability? I'm going to go out on a limb and say you didn't think that line of reasoning through.

In order to de-shittify Windows you would have to first de-shittify Microsoft, and that is never going to happen by cause Microsoft didn't get shittified, it has always been a shit company.

You should ask Gemini to read the summary for you next time so that you won't seem stupid.

You didn't get it. This is speeding tickets issued by fully automated systems. No court has ever seen this.

The solution is to stop treating it as a money-printing system. Punish people who speed in front of kindergardens, and leave people alone who at the city edge in the industrial zone on a Sunday evening didn't decelerate quite fast enough coming in from the highway.

I literally have no clue why you're saying that or who you think you're talking to.

Not at all. Because COMFORT is the key word here, not processing.

Which is often intentional.

We live in 2026. You can order an artillery or drone strike in Ukraine from a tablet and it'll be delivered within minutes. Data processing speeds are essentially instant for data sets not on LLM sizes. We have automation everywhere when it suits the right people. e.g. that speeding ticket you received for going 12 km/h over the limit? No human ever so much as saw that. Full automated process. Takes 6-12 weeks. Why? Because otherwise you might still remember how fast you were actually going or that you weren't even in that street on that day, and fight it. Same principle here: The slower the bureaucracy, the more people will just give up and go away. It's a self-regulating system to reduce the number of inquiries to the amount that the public servants can comfortably process.

If there is one thing I love about posts from you, it is that whenever I see a reply from you in my inbox I know I'm about to be entertained by a clown. LOL

I forgot to mention in my most recent post ... The extraordinary claim is the one made by you, to wit that the exponential improvement of AI models that has been going on for years and still continues has plateud and that you are the only one who knows about it. I'm always amazed at how truly stupid you actually are.

Remember' Chat-Gpt is just a word predictor! It doesn't think, and it will never be of any use! Only an idiot would be filled into thinking it does anything other than generate words! Thank you so much for trying to make a post to make me look foolish, and in doing so, unwittingly proving that you now admit that you have been clueless in every exchange we have had regarding AI. ROTFLMAO! Off you go now ...