IANAL, but there's a precedent in the anthropic ruling (page 11-ish), where the judge argues that using the books a-la-Google is fair use for training.

Anthropic got in trouble for using pirated books.

Google has scanned pretty much every book under the sun by buying it, digitizing, and destroying it and in that ruling that can be fairly used to train an LLM.

Google may settle just to get over it, but it has a decent fighting chance.

Imagine a beowulf cluster of those!!

I mean, there could be an associated bacteria that produces the toxin.

Morning Glory uses a fungus, maybe this fungus is using something else?

Oracle is extremely over-staffed.

I left when it was at ~120K employees (after Sun's acquisition), and the amount of people that did essentially nothing other than posturing was staggering (at least in the middleware division).

With proper organization and tooling it could have done the same or more with about half of the engineers, and that's *before* AI tooling.

The talent pools are bimodal, you have a small pool of extremely talented people that carry the weight of the org, and enormous swaths of seat warmers. There's a lot of politics, and headcount can be a source of power. Incentives were set so managers tried to hold as much headcount as possible to increase their leverage in the org.

we're still on Java 11 anyway

Siri has access to a lot of sensitive stuff, Apple is not willing to let any random app to access the same level of data without safeguards, the EU disagrees and says it should let anyone do it.

Would you trust Meta or OpenAI... heck or even Google with the same level of unrestricted access to device data?

The current EU legislation favors increased competition over personal privacy, Apple is not willing to make that tradeoff.
I think it's a reasonable solution just not to release it until both can be addressed.

The problem with String theory is that it doesn't predict anything. It just sits there being pretty.

At least try to explain something, like Dark Matter, and make a prediction we can test experimentally!!
Otherwise it's just a very convoluted form of masturbation.

Hopefully, Microsoft can keep the price at least somewhat lower than the $4,699 asking price for Nvidia's similarly specced DGX Spark box.

The DGX SparkBox has a ConnectX-7 NIC @ 200 Gbps. That alone is north of 1K of the price. It should be possible to get it at around 3K with 128GB of RAM with just 10G ethernet

The typical definition goes something like this:

Consciousness is the state of being awake, aware of one's surroundings, and experiencing subjective sensations, thoughts, and feelings

Think about a thermostat, it's awake, aware of it surrounding temperature, it "feels" that the it is too hot, which is unsettling, and causes it to signal the AC motor to turn on and suddenly feels ok, no more tension.

Consciousness is either: supernatural an ill defined or describes a simple feedback loop with some internal state.

We need a better concept.

Before release, you add a phase of vulnerability discovery, to find as much as you can with the latest and greatest models and fix those before release.
It makes sense to defuse the threat before it becomes an issue, otherwise your attackers will do it for you and won't tell you what they find.

I mean, Leia was mostly AI in the latest ones.

I mean, 10PB is not easy to store.

The moment you start charging for access, it's no longer open in the traditional sense.

It may be free as in freedom, but you cannot exercise that freedom unless you pay up?

We'll end up with a bunch of Mozillas trying to maximize ~shareholder~ community value.

SaaS is not going away, some may be replaced by AIaaS. At least for relatively simple things, but most companies paying for SaaS are paying for the support that comes with it.

You can definitely vibe code a lot of crap, but once you get into regulated environments where liability is really high (GDPR, HIPPA, FedRamp, etc.) it's risky to trust that a probabilistic interpretation of a prompt is compliant.

I mean, the picture and sound quality at home movie watching has gotten so much better that dealing with all the strangers in the cinema and all that hassle makes no sense.

In fact most HDR movies are way better in a large, modern TV than on a cinema screen.

And even mid-range sound bars with subwoofers can approximate the audio experience very well, algorithmic audio does magic with Atmos data.