I think TI calculators are not a source of distractions the way a phone or a laptop is

Liquid does amazing work, but I kinda feel like they are overtraining their models. 38T tokens seems like a lot for an 8B model

What's the downside? Don't they stop when they hit diminishing returns?

Wouldn't the model start overfitting at some point? Degrading generalization for accuracy on the training set.

You’d think so, but I haven’t seen it explicitly discussed in their papers, and nobody else that I know of trains on that many tokens

I like vibe coding for personal projects where the code is a throwaway, but I hate vibecoding at work when I later have to stake my reputation in validity of results, as well as maintain the resulting implementation.

Well the answer is easy, don't vibe code then?

I would love to, but in some companies these days that’s really not an option

It’s the healthcare for me.

And then it wants to edit some random upstream file that is not relevant to the task at hand and we should not edit it, so you tell it “and only edit the files affected by this commit”, and wait two more minutes.

And now it deletes a test, so you tell it “and don’t delete any tests”, and wait two more minutes.

And now it adds logic to disable the core functionality, so now you tell it “and don’t disable the core functionality”, and wait two more minutes.

Etc

After a refreshing walk it suddenly occurred to me - I had forgotten to add "make no mistakes".

They didn’t have showers, but you may recall Archimides shouted “Eureka!” after a famous bath time discovery

Would you happen to have more information about this online group? Would love to join it!

Sadly the (non-english) whatsapp group is no longer available. There wasn't really more to it; you had to practice circular breathing daily and post an update. But the teacher wanted to create membership schemes and make extra money, so people left.

He may not be a brilliant researcher, but he is a brilliant teacher. I am glad he is joining Anthropic so he can stay up to date with the next round of things that he will teach :)

It is a pity we don't hear more about the truly brilliant researchers.

All we hear is Altman, Musk, ...

> It is a pity we don't hear more about the truly brilliant researchers.

Reason? What is the value of that other than entertainment? And it's not in the interest of companies to make celebrities that then are poach targets (if they can avoid they would yes there are exceptions as noted elsewhere in this thread).

And if you did 'hear' (via articles) to what extent what was said even be correct vs. a writer just fluffing things up to the max.

Tech is not sports where you can actually see the superlatives and know that the person who praise is being lavished on actually won or threw or caught and so on. (Or even music where you can hear it and see the stadium that is packed with fans..)

Well whatever debt they took on to refill it would have been paid off quickly.

Buy at $60, start a war with Iran, sell at the new price, profit.

I guess the reason they didn’t do this is they thought Iran would fold quickly and oil would become even cheaper than $60

Nobody who was a decision maker on the war thought about any of this, at all, in any way.

There was no strategy, no thinking, no planning, nothing.

The US military has wargamed the Strait of Hormuz to hell, and all that was ignored, or dismissed, or not even considered to be relevant.

The full force of the US military is in the hands of a man who operates on whims based on what his reality TV instincts tell him will look good, and we are seeing the weakness of electing a reality TV star known for his capricious decision making and cruelty.

“Content”

Yes, memes and the like. Lots more than before. Entertainment is far better than it was before AI, when it was controlled by the big media companies.

Memes were never controlled by "big media".

But I'm sure some people LOVE that video generation is no longer "controlled" by there having to be a real thing to take videos of. You can just make shit up, and retreat even harder into bubbles. No need to interact with "the enemy" to even mock them, no need for source material to even twist or lie about; just generate whatever. There is a rift between "AI" and people, and one between those people vs. those who take having a society, culture and public spaces seriously.