So it's the age of AI. And this seems like a great new benchmark! Lots of text, structured but each item a separate "task". Each thing requiring its own new image + textual representation.
I copy + pasted the whole article (minus the few included images) and added this prompt in Gemini 3 Pro:
> Take each of the following and add an image representing the act being described. The image should be very basic. Think of signs in buildings - exit signs, bathroom door signs, no smoking signs, etc. That style of simplicity. Just simple, flat, elegant vector graphic lines for the chopsticks, hands, bowls, etc.
I think this is pretty dang good for a one-shot run. I also ran this through Claude Opus 4.6 Extended (doesn't generate images directly, so it made an HTML page and some vector icons). Not as good as Gemini IMO. See here if curious: https://claude.ai/public/artifacts/8b6589b3-4da4-4fd5-b862-c...
Anyone able to do this better with a different prompt or model (or both)?
You can buy them from various manufacturers that make them; you often get unsolicited mail from them as your name and address is on the patent filings.
Using Claude for code you use yourself or at your own company internally is one thing, but when you start injecting it into widely-shared projects like this (or, the linux kernel, or Debian, etc) there will always be a lingering feeling of the project being tainted.
Just my opinion, probably not a popular one. But I will be avoiding an upgrade to Node.js after 24.14 for a while if this is becoming an acceptable precedent.
I still think everyone is trying to run away from the copyright problems with AI, and suspect it's going to come back to bite them. Eventually. (No I'm not willing to bet on exactly when because I'm sure it'll be a lot longer than I'd like).
This page is a good example of when Typescript is both unnecessary and off-putting, based on the content's purpose.
The author is trying to demonstrate a problem and the proposed solution, using example code that the reader then needs to read, formulate a visual "structure" in their mind, and then apply that structure to the subsequent code.
Typescript is not in any way invalid here, and as a tool when programming something that will run in the real-world, can be invaluable.
However, when writing anything you want someone to deeply comprehend, you want to use the least amount of text needed to get your point across. Adding types don't serve any purpose here. The types used are generics, which tell you nothing specific, hence the name, and the names given to the functions and object properties are enough to convey what this code is doing.
Good luck finding a phone with a headphone jack anymore though :(
I love my wired headphones though. They support BT but I've used that maybe twice. Ever. Obviously was only because I was using my phone with them, which again don't have a port for the cord.
While AI is not a real human, brain, consciousness, soul ... it has evolved enough to "feel" like it is if you talk to it in certain ways.
I'm not sure how the law is supposed to handle something like this really. If a person is deliberately telling someone things in order to get them to hurt themselves, they're guilty of a crime (I would expect maybe third-degree murder/involuntary manslaughter possibly, depending on the evidence and intent, again, not a lawyer these are just guesses).
But when a system is given specific inputs and isn't trained not to give specific outputs, it's kind of hard to capture every case like this, no matter how many safe-guards and RI training is done, and even harder to punish someone specific for it.
Is it neglect? Or is there malicious intent involved? Google may be on trial for this (unless thrown out or settled), but every provider could potentially be targeted here if there is precedent set.
But if that happens, how are providers supposed to respond? The open models are "out there", a snapshot in time - there's no taking them back (they could be taken offline, but that's like condemning a TV show or a book - still going to be circulated somehow). Non-open models can try to help curb this sort of problem actively in new releases, but nothing is going to be perfect.
I hope something constructive comes from this rather than a simple finger pointing.
Maybe we can get away from natural language processing and go back to more structured inputs. Limit what can be said and how. I dunno, just writing what comes to mind at this point.
My companies makes potentially dangerous things like lawn mowers. We have a long set of training on how to handle safety issues that gets very complex. Our rules about safety issues is "design it out, then guard it out, and finally warn it out" - that is an ordered list so we cannot go to the next step until we take the previous as far as we can. (and every once in a while we [or a competitor] realize something new and have to revisit everything we sell for that new idea)
Courts will see these things for a while, but there have been enough examples of this type of thing that all AI vendors needs to either have some protection in their system. They can still say "we didn't think of this variation, and here is why it is different from what we have done before", but they can't tell the courts we had no idea people would do stupid things with AI - it is now well known.
I expect this type of thing to play out over many years in court. However I expect that any AI system that doesn't have protection against the common abuses like this that people do will get the owners fined - with fines increasing until they are either taken offline (because the owners can't afford to run them), or the problem fixed so it doesn't happen in the majority of cases.
Is the headline actually surprising to anyone? AI products that are currently live on a half dozen cloud providers are fueling thousands of people's various delusions right now.
No, the LLM itself is not a human, but the people running the LLM are real people and are culpable for the totally foreseeable outcomes of the tool they're selling.
The vendors will argue that the benefits that some people are gaining from access to those tools outweigh the harms that some other people like Jonathan (and like Joel, his father) are suffering. A benefit of saving a few seconds on an email and a harm of losing a life due to suicide are not equivalent. And sure, the open models are out there, but most users aren't running them locally: they're going through the cloud providers.
Same human responsibility chain applies to self-driving cars, BTW. If a Waymo obstructs an ambulance [1] then Tekedra Mawakana, Dmitri Dolgov, and the rest of the team should be considered to have collectively obstructed that ambulance.
I copy + pasted the whole article (minus the few included images) and added this prompt in Gemini 3 Pro:
> Take each of the following and add an image representing the act being described. The image should be very basic. Think of signs in buildings - exit signs, bathroom door signs, no smoking signs, etc. That style of simplicity. Just simple, flat, elegant vector graphic lines for the chopsticks, hands, bowls, etc.
Google Gemini output: https://gemini.google.com/share/11df1bc53e3d
I think this is pretty dang good for a one-shot run. I also ran this through Claude Opus 4.6 Extended (doesn't generate images directly, so it made an HTML page and some vector icons). Not as good as Gemini IMO. See here if curious: https://claude.ai/public/artifacts/8b6589b3-4da4-4fd5-b862-c...
Anyone able to do this better with a different prompt or model (or both)?
reply