At the time, it was assumed that genes and traits had a mostly one-to-one correspondence. Thanks to the human genome project we now know that it’s more often a many-to-many correspondence, which makes figuring out the relationships enormously more complex. But mapping the genome was still a critical step.

Edit: The analogous situation in neurology would be the correspondence between brain regions and cognitive functions—in the last decade or two we’ve found out that most functions involve many separate brain regions networked together in different patterns for different functions. Mapping this “connectome” is the equivalent of mapping the genome—but this Harvard/Google neuron-mapping project is at a much lower level, more akin to studying the physical structure of chromosomes.

Good point.

This cubic millimeter of tissue has allowed Harvard and Google researchers to produce the most detailed wiring diagram of the human brain that the world has ever seen. […] This helped researchers to identify the brain’s six cortical layers and white matter.

I thought the human cerebral cortex was two to four millimeters thick—how does a single cubic millimeter contain all six layers plus the underlying white matter?

It does suggest that tech users should reconsider their assumptions about the scope of the arbitration clauses in the user agreements they sign.

Humans instinctively do something analogous with natural language, using poetic forms like rhyme, meter, and alliteration. (For example, the speeches from Shakespeare’s plays are immediately detectable because they’re in iambic pentameter.)

Imagine you lacked the natural human ability to detect verse, making poetry indistinguishable from prose. As far as you could tell, it would be like an invisible watermark that only specialists could detect. LLMs can use a similar approach, making up their own patterns that are opaque to humans but detectable to themselves.

The obvious missing element is another AI on Sydney’s end to summarize all the fan mail into a one-number sentiment score. At that point we can eliminate both the AIs and the mental effort, and just send each other single numbers via an ad-sponsored Google service.

AFib patients using wearable devices are more likely to engage in high rates of symptom monitoring and experience anxiety than non-users

Well no shit—how can non-users engage in high rates of symptom monitoring if they don’t have symptom monitors?

As Mary Anne Franks, a George Washington University law professor and a leading advocate for strict anti-deepfake rules, told WIRED in an email, “The obvious flaw in the ‘We already have laws to deal with this’ argument is that if this were true, we wouldn’t be witnessing an explosion of this abuse with no corresponding increase in the filing of criminal charges.”

We’re certainly witnessing an explosion of media coverage of abusive deepfakes, as with coverage of everything else AI-related. But if there’s no increase in criminal cases, what’s the evidence that the “explosion” is more than that?

For those not wanting to read the article, note that they revealed (to employees) a progress framework, not any actual progress.

The framework is just a five-tiered classification of potential future AIs: Chatbots (1); Reasoners (2); Agents (3); Innovators (4); and Organizations (5). They characterize their current progress as near level 2, but there’s no indication of recent progress that would be newsworthy of its own accord.

“We will now remove content that targets ‘Zionists’ with dehumanizing comparisons, calls for harm, or denials of existence on the basis that ‘Zionist’ in those instances often appears to be a proxy for Jewish or Israeli people,” Meta said in a blog post.

So dehumanization and calls for harm are fine, as long as the target isn’t a proxy for Jews or Israelis?

The fallacy is imagining that “lawfulness” is an attribute that can be reliably detected on an implementation level.

I guess having a lot of unhappy customers implies that a lot of people previously purchased the product.

All ai projects should be forced to show the entirety of their training data.

Agreed—but note that in this case the information was only discovered because the organizations involved (Common Crawl and LAION) do show their data. We should assume that proprietary data sets have similar issues—but this case should be seen as an opportunity to improve one of the rare open data sets, not to penalize its openness and further entrench proprietary sources.

Michaels, 79, told Vanity Fair in an interview published Wednesday that he was initially “very skeptical” of the proposal from NBCUniversal executives — until he heard the AI-generated version of his speaking voice, which is capable of greeting viewers by name.

Was this a phone interview, by any chance?

I propose detecting atmospheric anomalies induced by their infinite improbability drives.

While the labels give retailers the ability to increase prices suddenly, Gallino doubts companies like Walmart will take advantage of the technology in that way. “To be honest, I don’t think that’s the underlying main driver of this,” Gallino said. “These are companies that tend to have a long-term relationship with their customers and I think the risk of frustrating them could be too risky, so I would be surprised if they try to do that.”

How to tell if an academic doesn’t get out enough.

An organoid is not a single cell—each one can have thousands of neurons, depending on the size.

Here’s a video that starts with a good general overview of brain organoids:

https://www.youtube.com/watch?v=x1Pg56WWm5U

Can it do backpropagation?

this data is not the world, but discourse about the world

To be fair, the things most people talk about are things they’ve read or heard of, not their own direct personal experiences. We’ve all been putting our faith in the accuracy of this “discourse about the world”, long before LLMs came along.