They have trouble adding two numbers accurately though

Why are they expected to?

"Make sure quotes in your article are things the subject actually said to you" is not something that should need a "learning moment".

That absolutely should be career-ending for a journalist, apology or no

Not "career-ending" but definitely back a few paces.

This wasn't outright fabrication, it was a sloppy editorial workflow that resulted in hallucinations being published as fast (which is absolutely going to be get more and more common unless newsrooms develop specific guards against it).

How is it not "outright fabrication"? Quotes were attributed to a subject that he didn't say. That's fabrication.

Genuine question re #1: does your text editor not already do that?

Without ai my text editor auto completes letters into existing identifiers or adds a closing brace

With ai it add several lines of code at once as soon as it thinks it recognizes a common pattern.

It’s not perfect and it can get in the way but it’s amazing when it guesses right and spits out the next 3-4 lines I would have typed

Personally I find this workflow is jarring. I get into flow typing code and then the AI autocompletes the next four lines on a tab input. Now my flow is screeching to a halt because I have to switch from flow mode to review mode to make sure it actually autocompleted what I wanted

I'm trying to think of a text editor that doesn't support customizable snippets and templates, and failing

Those two things aren’t the same at all, they’re so different it’s hard to believe you’re not being intentionally obtuse.

I mean, in the sense that customizable snippets are more flexible and can cover a wider variety of use-cases, I guess?

Text editors/IDEs have simple autocomplete and the ability to do some expansion, e.g. a for loop with placeholders to fill in. Those work and are still useful.

JetBrains also has local line-based LLM models for various languages.

With the LLM-based autocomplete it a) generally autocompletes more code at once, and b) will often pick up on patterns in the existing code. E.g. if you have a similar method, list of print/string buffer write statements, or other repetitive code in the file it will often use that as a model for the generated code.

That sure sounds like you're describing customizable snippets, which AFAIK every major editor supports?

Sitting here on the sidelines having never configured snippets or macros or any of that in any of my editors, which I could have done like 30 years ago but never bothered in all this time, doing quizzical-dog look at all these people thrilled about LLMs.

I guess they might finally get me to use those things since they take the “configuring” and “remembering shortcuts” part out, but so much of this doesn’t look new at all. Super old, actually.

In my objective opinion, almost all AI uses cases (coding or otherwise), are just because of people's extreme laziness in spending a little time setting up some "automated" workflow, be it canned templates or whatever. The non-AI approach has the added benefit of being precise!

Customizable snipping is a feature editors support (which I mentioned as they are related/similar to what the AI is doing), but is different to the AI autocomplete behaviour.

If I have a JSON structure, I can paste that into the file as a comment, e.g.:

    # {"foo": 1, "bar": "test", "baz"}
    @dataclass
    class FooBar:
        foo:

and the AI will/can autocomplete/generate that to:

    @dataclass
    class FooBar:
        foo: int
        bar: str
        baz: int

using the JSON example. Then if you type:

        def __str__(self):

the AI could then contextually generate, e.g.:

    return f'Foo(foo={self.foo}, bar={self.bar}, baz={self.baz})'

Or if you have a SQLAlchemy model:

    class Foo(Base):
        __tablename__ = 'foos'

    bar_id: Mapped[int | None] = mapped_column(ForeignKey('bars.id'), default=None)

typing `bar:` the AI can autocomplete:

    bar: Mapped[Optional['Bar']] = relationship()

picking up that you have a `Bar` class in the file. Especially if you have other similar id/object definitions in the file.

It's a DWIM button.

The JetBrains local autocomplete is hilarious but occasionally useful. I find it really hit and miss in terms of when it will decide to autocomplete and whether it will exhastively complete all elements, miss some out or get itself into a loop over several.

The out-of-the-box stuff is supposed to be kind of stupid. Are you guys really not editing your own snippets and shortcuts? Have people really been typing out "def do_something(foo, bar, baz)\n\t" manually?

Is it though? Really? I'm still waiting for even one of my vendors, commercial or open source, to actually speed up their release cadence.

VS Code Insiders is releasing 3 times a day.

Used to be at most once.

That's actually a great point: judging by the dev team's commits at work there's an unprecedented amount of code being committed but it's not actually making it into releases any faster. Maybe the same thing is happening at my various vendors, but then that kind of argues against the idea that Everything Has Just Changed.

Plenty of commits link to mailing list discussions about the proposed change, maybe something like that, with an archive of LLM sessions?

Huge and great news. Sri Lanka is hoping to get certification later this year too.

I don't know that OpenAI specifically is the weak link but this definitely adds to the argument that the entire sector is a wash with the same three or four companies passing around the same $50B over and over. OpenAI is just the link that seems most likely to break first.

I've subscribed to a few AI tools for the last 3 years now. I'm someone who almost never subscribes to anything.

I'm sure that $50b has my money in there somewhere.

Your subscription fees didn't fully pay for the amortization on the chips you used for the inference.

Inference is profitable. Gross margins are in the 50%+ apparently. This is coming from Anthropic CEO.

I'd say that too if I were an AI CEO looking for funding

Yep same, I'd sooner starve than cut my Anthropic sub

If tomorrow Claude pricing changes and they start charging real API costs like 2000+ USD, and there is another service: "NotReallyClaude" that is a bit less good but 200 USD, then what would you do ?

Man, they really got good at hitting that dopamine button, huh?

It's not "continue" buying as much as this is NVIDIA fronting the money for (most of) the hardware OpenAI has already ordered from them. It's like borrowing rent money from your drug dealer.

Great analogy. ;-)

Doubt Jensen sees himself as a “dealer” but considering the vendor lock-in and margins, he pretty much is the Tony Montana of Ai Chips.

It’s nuts that this type of financing is legal.

It's like credit cards loaning money to people who are unemployed and will default on payments. It's a risky business that is legal and can be very profitable, but may also be disastrous in the future.

>It’s nuts that this type of financing is legal.

You need people to burn in house fires for regulation to require extinguishers.

We're going to be the next generation’s cautionary tale.

I don't see the problem as long as materially significant transactions by publicly traded companies are properly disclosed to investors. If someone loses money by buying NVDA then they have only themselves to blame.

This is Jeremy Irons' argument in "Margin Call" too. But most people were unhappy with the secular result.

Tuld wasn't wrong. There will always be financial bubbles and misallocation of capital. It can't be prevented, and even trying to prevent it would involve intrusive government overreach that would make most people even more unhappy. Investors who want safety are free to buy Treasuries.

It is legal because Jensen isn't selling drugs, payday loans are legal too!

It’s legal because both sides have armies of lawyers and are voluntarily entering into contracts where each party gets consideration.

How someone can compare the above situation to a person getting a payday loan to put a roof over their head or food on their plate is beyond me.

The “it’s like <insert wild and inappropriate analogy to stoke emotion>” is a tired trope.

Come on, calling a round of vendor financing (which is what the NVIDIA money is) "funding" is eggregiously misleading. The only new money entering the sector from this is SoftBank's stake.

They might have dressed up the wording, but the details are all there for anyone who wants to objectively look at the deal. It is a group of two executives making a non coerced deal and disclosing the required information to investors.

Might be a stupid gamble, but it's not akin to a loan shark shaking down a hungry, cold person for life's essentials.

Conversely it’s equity for an in-kind investment. Dave Choe taking the Facebook shares writ large.

I'm annoyed enough by coworkers asking "is the server down?" that I try not to do the equivalent to other people at their jobs, particularly doctors.