tomohelix

The author basically chose some of the output based on set criteria. I think this can eventually be automated and embedded into the protein language model the same way ChatGPT now has guardrails and specific ways to answer questions, instead of following up with the most likely sentence, e.g.: asking it what is the capital of france get an output of another question about what is the capital of germany.

So I am tired and only make a few posts every few days now on HN. I am sure while my activity has dropped, the bots are getting more active so nothing is lost. Maybe some quality and how the traffic share look like, but I don't know.

Only a miniscule amount of protein has verified structure and function. While an HMM can predict their functions pretty well based on homology, it can often be wrong. Having that kind of noise in the training data can degrade the performance of these protein models. Perhaps we need better curated dataset before we can scale these up.

You are probably right that lysozyme is an easy target and may have large sequence variety between homologs so saying "very different" for 30-40% is not correct. But that is only in the context of biology and protein structure and function. This is an LLM trained on primary sequences only. It doesn't know anything about the folds or domains or functional sites (unless I am wrong and those are part of the metadata fed to it during training). Yet it did learn enough to generalize to the point that even with only 30-40% identity, it still produces soluble proteins with the same function. I am sure you know that at 40% differences, one protein can be in an entirely different superfamily from another. So it is still an impressively low identity score.

Also, I think it is more appropriate to compare the amino acids to things like the alphabets than vocabs. Domains would probably be an equivalent to LLaMa vocab.

Well, technically it is out in 2023 but sure, you can argue it was probably in the prepub state since 2022...

The interesting thing about this model is that it also exhibit emergent capabilities. It was trained only on raw sequences but somehow managed to capture information about functionality and solubility of the folded proteins, and then implemented that in the generated sequences.

Amino acid sequences are just a bunch of jumbled words if you compared them to English. It usually has to go through folding to form proper "sentences" with meanings. I guess you can compare this to "grammar". This is probably the model managed to learn protein grammar purely by brute force. Now if only we can get a model in the range of 100bln parameters...

Or, CEOs receive 50k/yr compensation, but the company will approve a certain discretionary spending amount every year to cover living expenses for the CEO. This may or may not include dining at the most expensive restaurants around, free travel to anywhere with any arbitrary amount of number of people, house purchases, luxury spending like a yacht... No, these are not regular compensation. We are simply subsidizing the living expenses. There is no concrete amount involved. No sir.

Or pick any other number of convoluted schemes to mask "compensation". Best way to do this is the easiest way, raise the corporate tax directly.

An LLM is so different and alien we would literally incapable of imagining their thought process even if they are pummeling us with evidences of their intelligence. It is just how it is. So I think it is futile and even foolish to try to quantify or create criteria for "intelligence" when discussing AI. Let it be and accept it for what it is: something we humans have created that can do lots of stuff.

They are calling fraud because OAI was mixing nonprofit and capped profit to avoid taxes and the likes. They also made claims about unfair representations coming from ChatGPT and the executive composition of OAI, which apparently looks like insider trading and information manipulation. They alleged that OAI participated in election interference during 2020. They also said OAI was recklessly making advances in disruptive tech faster than legislators can make laws for them.

Basically, a prime example of a vulture coming in to sue anything that looks like it can give them some money. This is part of the reason why every company need a team of lawyers to deal with this perversion of justice.

You are confusing the parent company with the publishing part for that 19%, which is the main subject here. You are uninformed.

Look under the Scientific, Technical & Medical financial review. That is their scientific publishing arm.

https://www.relx.com/~/media/Files/R/RELX-Group/documents/pr...

Most journal reviewers and editors are academics who do the work as public service, i.e. free. Even top spots like editor in chief only get an honorarium of maybe 10k a year. Considering these people are the top of their field and can easily command 6-7 figures salary at minimum, that amount is nothing. Only very few journals have full time editors and even those end up relying on the volunteers to vet and review articles.

A full time professor is expected to be an editor for some journals. And these academics make up the bulk of "journal editors". But they would not list it as their jobs. I suspect that is why if you look for "editor jobs" you will get lobsided salary ranges that doesn't reflect reality.

All those "publisher" middlemen can go eat dirt and I would be happy to shovel more dirt on them until they are 6 feet under. They do no work, take all the profit, and obstruct research and the spread of knowledge. Those are some of the worst scums of capitalism.

We used to think 100mph was the fastest speed we could achieve even with machine (back when trains were the cutting edge). New tech will change things quickly, all we need is one breakthrough and things will fall into place.

Also, faster is understated. You are talking about the speed difference between chemical gradients and the speed of light.

This would be like if internet social media came out and the laws tried to control it using rules similar to physical books and newspapers. They won't be effective and will just create a hostile environment for development of these techs where these laws are in effect. Meanwhile, those who don't care about doing these control freak things will develop the tech and dominate the new sector.

The EU need to sit down and really think about how to control AI tech properly instead of passing kneejerk and lazy regulations. It isn't easy to make a legislation framework for something entirely new. But it can be done instead of ruining the whole thing.

This isn't like Ukraine where there was an implicit agreement that nobody will send actual troops into Ukraine to fight its war. If Taiwan is invaded, the US will send its own troops. And if the US is involved, the EU is likely involved too. China is big but not enough to challenge the US militarily. Not yet.

It would take an absolute madman to make China go to war just for Taiwan.

Uh, isn't that how it should always be? If the output isn't conditional on the input, then it is basically noise or worthless. How the conditionals are setup are the key but not really sure how does this relate to the point you are making.

But still, we already had lots of evidence how some animals like cetacean and primate are potentially sentient. Many people just don't want to admit it and we can't convince them that human isn't anything special. In the end this may just be another piece of proof on a pile that get ignored.

We are so close to something amazing, and scary.

You are also free to respond to the other comment I made to your post and not jumping to another one. These remarks are not specifically for you. Go take a look at the account I replied to above yours to see why I said that.

I tried giving them benefits of the doubt before, it never went anywhere. Maybe here it is different. I guess I can try again.

The context in this specific part is about the retaliations between countries. It has expanded beyond just one company when you start talking about entire economies fighting each other.

Besides, I already gave you the US hardware that China banned.

Also, you are making a bad faith argument. The topic was in "product". You specifically tried to gate it to "hardware". China is the producer of most hardware in the world. They aren't going to ban stuff that was made by them. Software product on the other hand, aren't really their forte until recently and that they banned plenty.

We have cover texts, images, videos. I wonder what is going on with audio. Once all of these get merged into one combined model, I honestly don't see any differences between a random person on the net and an AI. The only ways we can interact with each other through the internet are those 4 things.

Not all enzymes are stereospecific though. Some of them can take any configuration of a chemical and convert them to other type of isomer. But these are rare and usually an exception, not the norm. Still, it makes it harder to figure out why chirality is so important in life and not so much in some other parts of life.

Maybe your explanation about handedness makes sense. But then it would fail at explaining ambidextrous people since they are definitely better and yet still so rare. Life is really difficult to figure out.

Don't worry, I am not an atom-ist or a chiral-ist. All sides are equal! (except the ambidextrous master race :D )

Last time I read about this, I think some physicist said something about the rotation of elementary particles and radiation stuff that are way over my head.