Don’t forget: amethyst on the windowsill.

Possible, but seems unlikely. Unless there’s some verified record, the mere fact he may have taken a valid test is very weak evidence that his claimed scores are accurate and not exaggerated. 

Just leaving this here on the off chance this comment counts and I can claim the bitcoin. 

This comes across as total nonsense. 

Perhaps the jailbreaking issue is easier to grasp.

“All modern AIs can be jailbroken to act dangerously.

Nobody knows how to solve this.

When we get really powerful AI, bad guys can use it to kill millions, maybe billions”

I am willing to bet you at 5:1 odds in your favour that OpenAI does not lose more than 50% of its valuation within 1 year from today for up to $1000 of my own money.

Grok 4 doesn’t appear to be a meaningful improvement over other SOTA models. Minor increases in benchmarks are likely the result of Goodharting.  

I expect that GPT 5 will be similar, and if it is, this gives greater credence to diminishing returns on RL & compute.  

It appears the only way we will see continued exponential progress is with a steady stream of new paradigms like reasoning models. However, reasoning models are a rather self-suggesting and low-hanging fruit, and new needle-moving ideas will become increasingly hard to come by.

As a result, I’m increasingly bearish on AGI within 5-10 years, especially as a result of merely scaling within the current paradigm.

There's a good overview of his views expressed in this manifold thread.

Basically:

• Caloric restriction works, however it impedes his productivity ("ability to think").
• Exercise isn't effective in promoting weight loss or reducing weight gain due to compensatory metabolic throttling during non-exercise times
• His fat metabolism is poor, because his fat cells are inclined to leach glucose and triglycerides from his bloodstream to sustain themselves rather than be net contributors, and the effect is that muscle loss makes up the difference, leading to unfavourable implications for body composition, energy, and overall health. In essence "good genetics" for fat loss = "Fat cells that are readily and efficiently broken down for energy", whereas "bad genetics" for fat loss = "fat cells that are resistant to being used as energy".
   

It’s starting to really feel like we’re in the process of AI improvement fizzling out and companies are merely disguising this with elaborate products. 

God I hope you're right.

I found myself chuckling at all of them because of this exact reason. Has GPT stumbled on an ingenious form of meta humour?