imperfect_light

120 biolabs in over 30 countries

And yet the only ones listed in detail are the ones in Ukraine, huh.

You should ask BigCo if they're only providing goods/services to their customers "for the money."

He did so well with Theranos

Now we're seeing more colon cancer in young people. All of the discussion is around factors that change slowly (diet, alcohol, exercise), no one I've heard has pointed out that a pathogen would explain a quick change.

Have always been fascinated with Paul Ewald's arguments (as laid out in his books The New Germ Theory of Disease and Evolution of Infectious Disease) that most chronic illnesses are due to pathogens.

This seems similar to the "let's just rebuild from scratch" impulse that has been tried so many times on very large complicated systems and often, although not always, fails.

I'm not arguing that layoffs lead to a death spiral.

I'm arguing that if you're on a downward trajectory and your primary strategy is layoffs and cutting expenses, rather than investing, you're not going to recover.

Anyone have examples of companies that were in trouble and cut employees and spending heavily and that led them back to success?

My personal experience with working for such companies is that it leads to a death spiral, but I recognize my sample size is small.

My question is less about PDFs and more about the notion that all the facts needed for the RAG are in documents. In my experience just a fraction of the questions that might be useful exist in a document somewhere. There must be a variation of RAGs that are pulling not from documents, but from databases using some semantic model.

The emphasis on PDFs for RAG seems like something out of the 1990s. Are there any good frameworks for using RAG if your company doesn't go around creating documents left and right?

After all, the documents/emails/presentations will cover the most common use cases. But we have databases that have all the questions the RAG might be asked, far more answers than that which live in documents.

How do you know you're actually getting what you bought?

Quantum entanglement

Police in my city go to pretty much every protest and take video and photos (with big zoom lens) from the roofs of nearby buildings. It's always surprising to me that more people don't notice.

You're conflating outright fraud with a difference of scientific opinion?

Where are these numbers from? How does 5 million housing units from 2020-2023 (this was published in 2024) translate into "90K housing units"

The United States has added almost 5 million housing units since 2020, most heavily in the South and most of them single-family homes, making a housing shortage look conquerable in much of the nation.

https://washingtonstatestandard.com/2024/05/16/housing-boom-....

If you make them feel good, they’ll say yes.

You could really generalize and say this is about relationships in general. If people leave interactions with you feeling good about themself, it will be a relationship they want to continue and grow.

And the ratio of good interactions to bad needs to be at least 5X (at least that's what the Gottman Institute found with romantic relationships)

You think they all magically figured out they were overstaffed at the same time? It's 100% herd mentality. They're cutting because everyone else is cutting, just like they went on hiring sprees because everyone else was doing the same.

It's easy to measure short-term impact (we cut a bunch of people, we're saving money, we're more profitable) but it's very hard to measure the medium to long term impact of these cuts.

Note I'm not arguing these cuts are the wrong strategy, I'm arguing they have absolutely no clue.

Reminds me of the famous Reagan "welfare queen" story about someone showing up in Cadillac to use their food stamps. Did it happen, probably. Is it widespread, it is representative of most people on food stamps. Of course not.

Same situation here. Of course it's happened, some people have taken advantage of remote work. So what's the manager's excuse for not catching this?

But SQL is an old language with significant design problems, making it difficult to learn, difficult to use, and difficult to extend.

It's difficult to learn and use? Compared to what? The "difficult to extend" might be accurate, but I really question the initial premise here.

What are you basing this assertion on? It was closed because of the pandemic. When that was over Allen had been dead for several years, his instructions had been to sell everything and give it to charity so that's what his sister has been doing. No one attempted to reopen with alternative benefactors or see if visitors would come.

I was there right before the pandemic and there were a decent number of visitors (given that it was a weekday afternoon).

If the museum was open I'm sure it would have plenty of visitors and support. Can't visit a museum that isn't open.

But the fact is that most museums rely, in least in part, on grants and benefactors in addition to visitors.

It's not like there's a shortage of money for art museums or natural history museums, but we do seem to be closing museums (this one and previously the Boston one) detailing the history of one of the biggest innovations in human history.

When you say "world needs" you mean sports teams and personal space travel? It's safe to say all of them have plenty of money to save this small museum and still fund their hobbies and causes.

The dude was sick on and off for years, could have had an army of lawyers setting up these places to continue after him, and didn't bother. Don't blame his sister who was left this mess.

but likelihood that any of your potential heirs is also interested in it

That's not the issue, he set up the parts of his empire he cared about so they'd live on (Allen Institute for example) and instructed his sister to sell the rest. She could have gone against his instructions, but the point is he clearly he didn't care whether the Cinerama or the Living Computer Museum continued on. That's on him, not his heirs.

The fact that no other Seattle-adjacent computer billionaires like Gates, Bezos, Simonyi, Ballmer has offered to continue it just shows the generally low quality of people that have gotten rich from tech.

Test your website properly before popularizing it so people don't get errors like "Failed to execute 'json' on 'Response': body stream already read"

He's said it pretty clearly, an LLM could be part of the solution in combination with program synthesis, but an LLM alone won't achieve AGI.

You're reading a whole lot into a tweet, in his interview with Dwarkesh Patel he says, about 20 different times, that scaling LLMs (as they are currently conceived) won't lead to AGI.

The stretch was in reference to comparing Chollet to Einstein. Chollet clearly understands LLMs (and transformers and deep learning), he simply doesn't believe they are sufficient for AGI.

A bit of a stretch given that Chollet is a researcher in deep learning and transformers and his criticism is that memorization (training LLMs on lots and lots of problems) doesn't equate to AGI.

Did you listen to what Chollet said? How much of LLM improvements are due to enlarging the training sets to cover more problems and how much is due to any emergent properties?