trollbridge

History has proven that the microkernels are not necessarily more secure and have their own unique set of problems (see: macOS).

Quite the opposite, particularly when the biggest vendors of coding agents insist on not allowing their models to be used to check the code they generate for security issues.

I suppose you've got a point; if the middle class of Americans constantly have their careers destroyed, eventually we'll return to a third world sort of existence where childcare is provided by extended relatives.

However, most people in America can't rely on grandparents for childcare. One reason is we have children much older than people in the third world do - if two generations have children at age 38, it's not very reasonable to be expecting a 76 year old to provide child care.

No thanks. The last thing I want is the government with even more motivation to shove AI into everything in order to make it "profitable".

Many of these jobs are things that improve quality of life for everyone else, too, like street beautification programmes. They do this alongside coming up with robotic means to also sweep streets, but yes, they find ways to hire people to just "make this place look nicer".

Well, first these AI companies would have to turn a profit.

It used to be you could just buy (for example) jeans made in America for $20. I bought such a pair back in the late 1990s. There was no need to be busy trying to "support the traditional garment industry".

Nowadays you'll be spending $40 (same price after adjusting for inflation) for something definitely not made in America, and probably of inferior quality too. Since the domestic garment industry has shrunk so much, the supply chains an economies of scale don't exist anymore to buy domestically-made garments of decent quality at those prices.

Well, a lot of them retrained into call-centre type of jobs.

Then the call centre jobs offshored (and now are being replaced by AI).

I'm interested in how you expect someone to attend school/retrain without childcare. You can't generally bring your kids to class with you.

People don't get a degree in "generic doing things people want". They generally get one focused on a particular field.

If someone's degree is in computer science, they're not going to be a good fit to be, say, an accountant or a nuclear engineer.

The IA/A was general purpose.

S3’s stuff was no more “3D” than the Image Adapter/A was.

You should look at the IA/A - it had its own C like compiler, CPU, etc which did things reminiscent of a modern GPU or SIMD.

Run a combine and you’ll see.

Similar to problem to how 100% autonomous vehicles don’t exist, yet. There are too many edge cases.

Get to 99% first.

Yep.

Behavioural analysis of how the user uses the site to identify real people.

Then, textual analysis to detect when they’re promoting something specific. This is not hard to detect across the set of all active posters.

In the electrician subreddits, it’s painfully obvious whenever something is a bot, since… nobody ever really has anything terribly positive to say about any new brands. There are a few established brands with respect like Fluek, Knipex, and Klein. But it’s mostly a community of haters.

Periodically an account shows up promoting something asinine and it sticks out like a sore thumb.

There are farms near Sacramento airport in a “reversed irrigation district.

They have to pump water OUT of their fields and pay for the privilege of doing so as opposed to pumping it in from an irrigation canal.

Nothern California has no shortage of water whatsoever, and indeed it’s a regular news item whether or not to divert a few million extra gallows into the ocean to sustain some variety of fish that likes to swim around in the delta.

I have no idea where it came from but my current setup does it nicely.

It turns out the moat is “writing drivers that work”; Nvidia drivers simply work, and Intel’s are poor quality. So if I want stuff that works I need to buy Nvidia gear.

IBM shipped the first PC GPU (the Image Adapter/A, 1989) following on the first PC 2D accelerator (the 8514/A, 1987). There was zero benefit to being first. 3dfx had the first mass market, cheap GPU in 1996.

ATI (AMD) in 1989 copied the unpatentable parts of the 8514/A, improved it, and went on to dominate 2D accelerators.

nVidia’s first 3D card was a complete flop. They did not achieve success for years.

Nvidia is in the right place at the right time.

Intel could have beat the pants off Nvidia a long time ago with Arc if they’d bothered to ship usable drivers. But they refuse to, and simply can’t figure it out, so Arc cards remain cheap because they’re so #%£€ing hard to get working well, and everyone is nervous they’ll lay off the driver team again.

No.

Okay. About 50 million households in America earning under $50k.

8 billion is enough to give each of them $160. Great, all of Anthropic’s spending is enough to give each family a trip to Texas Roadhouse including dessert.

Because it is, because every other shareholder gets diluted.

apparently very few people - I know tons of people who use it and almost nobody who pays for it, particularly outside of tech

Well, agents are intentionally trained with RLHF to solve captchas that say “prove you’re human”.

It indeed is and points to serious problems in the criminal justice system that destroy the argument of whoever is using it as a threat

Well:

1. Park your Python or whatever code inside a VM with no connectivity to anything (other than to accept inbound ssh) and with a canned set of PyPi etc packages available for it to use

2. Park hypervisor for that in a machine with no connectivity except thru a firewall that only admits the relevant ssh traffic.

3. Make sure to use ssh clients that can’t be exploited by a remote server. If this is too challenging, then use telnet or rlogin instead.

4. You can extend this to eg allow outbound calls to an LLM. Alternatively, place the GPU hardware and LLM weights inside the hypervisor machine.

Congratulations, now nothing can escape except what you allow to from the output from rsh.

It’s beyond absurdity at this point, but the marketing plan of “describe our poor sandbox design as powerful agent capabilities” clearly works.

Reality is more mundane.

For example, it had it instructed not to open PRs and sandboxed to not be able to use gh to do so. Astra 6 has a really strong drive to open PRs, so it copied a set of browser cookies out and then instrumented opening a PR that way.

A similar example is asking it a question if we can do X, and then it will go and actually implement X.