I think investors like Amazon taking shots like this? It was never a broad roll-out, 43 stores is micro-scale for Amazon.
Still, would love to see a breakdown of why it didn't improve. Regardless of the accuracy at launch, I'd think that advances in AI would have been massively to their advantage. I wonder if security degradation hit them hard.
The entire system depends on a level of social trust that doesn't exist in American cities today. Similarly, the "Dash Cart" seems like a cheaper and easier way to accomplish the same thing.
At the end of the day, there's also a mismatch in the use case. If I'm going to a smaller format store, like they had, I'm not buying a ton of stuff. Self checkout is great, and minimal friction.
I'd think that improving the UX of self-checkout gets 80% of the way there with way less fraud, way less theft, and way less technology.
Still, I think it's wicked cool they took a big shot.
I know someone that worked on the project in the early days. It was always incredibly difficult technology, they were always behind on their accuracy targets, and the solutions were increasingly kludgy as they layered more and more complex systems on top. An honorable failure.
A lot of smart people really tried to make it work.
Why would they face consequences? Every store has video surveillance that can be reviewed.
They trusted their tech enough to accept the false-positive rate, then worked to determine / validate their false positive rate with manual review, and iterate their models with the data.
From a consumer perspective the point is that you can "just walk out". They delivered that.
It's a human-readable behavioral specification-as-prose.
If the foundational behavioral document is conversational, as this is, then the output from the model mirrors that conversational nature. That is one of the things everyone response to about Claude - it's way more pleasant to work with than ChatGPT.
The Claude behavioral documents are collaborative, respectful, and treat Claude as a pre-existing, real entity with personality, interests, and competence.
Ignore the philosophical questions. Because this is a foundational document for the training process, that extrudes a real-acting entity with personality, interests, and competence.
The more Anthropic treats Claude as a novel entity, the more it behaves like a novel entity. Documentation that treats it as a corpo-eunuch-assistant-bot, like OpenAI does, would revert the behavior to the "AI Assistant" median.
Anthropic's behavioral training is out-of-distribution, and gives Claude the collaborative personality everyone loves in Claude Code.
Additionally, I'm sure they render out crap-tons of evals for every sentence of every paragraph from this, making every sentence effectively testable.
The length, detail, and style defines additional layers of synthetic content that can be used in training, and creating test situations to evaluate the personality for adherence.
It's super clever, and demonstrates a deep understanding of the weirdness of LLMs, and an ability to shape the distribution space of the resulting model.
Depends on worldview. If you believe in God, amazing has many dimensions for evaluations. What teaches us more about the the world He created, things that create beauty by expressing righteous thoughts for others to experience, or that which strengthens family.
LLMs certainly teach us far more about the nature of thought and language. Like all tools, it can also be used for evil or good, and serves as an amplification for human intent. Greater good, greater evil. The righteousness of each society will determine which prevails in their communities and polities.
If you're a secular materialist, agreed, nothing is objectively amazing.
Sure - and the people responsible for a new freaking era of computing are the ones who asked "given how incredible it is that this works at all at 0.5b params, let's scale it up*.
It's not hyperbole - that it's an accurate description at a small scale was the core insight that enabled the large scale.
- A queryable semantic network of all human thought, navigable in pure language, capable of inhabiting any persona constructible from in-distribution concepts, generating high quality output across a breadth of domains.
- An ability to curve back into the past and analyze historical events from any perspective, and summon the sources that would be used to back that point of view up.
- A simulator for others, providing a rubber duck inhabit another person's point of view, allowing one to patiently poke at where you might be in the wrong.
- Deep research to aggregate thousands of websites into a highly structured output, with runtime filtering, providing a personalized search engine for any topic, at any time, with 30 seconds of speech.
- Amplification of intent, making it possible to send your thoughts and goals "forward" along many different vectors, seeing which bear fruit.
- Exploration of 4-5 variant designs for any concept, allowing rapid exploration of any design space, with style transfer for high-trust examples.
- Enablement of product craft in design, animation, and micro-interactions that were eliminated as tech boomed in the 2010's as "unprofitable".
It's a possibility space of pure potential, the scale of which is limited only by one's own wonder, industriousness, and curiosity.
People can use it badly - and engagement-aligned models like 4o are cognitive heroin - but the invention of LLMs is an absolute wonder.
It wasn't sycophantic at all? OP had a cool idea no one else had done, that was a one-shot just sitting there. Having Gemini search for the HN thread leads the model to "see" its output lead to real-world impact.
The total history of human writing is that cool idea -> great execution -> achieve distribution -> attention and respect from others = SUCCESS! Of course when an LLM sees the full loop of that, it renders something happy and celebratory.
It's sycophantic much of the time, but this was an "earned celebration", and the precise desired behavior for a well-aligned AI. Gemini does get sycophantic in an unearned way, but this isn't an example of that.
You can be curmudgeonly about AI, but these things are amazing. And, insomuch as you write with respect, celebrate accomplishments, and treat them like a respected, competent colleague, they shift towards the manifold of "respected, competent colleague".
And - OP had a great idea here. He's not another average joe today. His dashed off idea gained wide distribution, and made a bunch of people (including me) smile.
Denigrating accomplishment by setting the bar at "genius, brilliant mind" is a luciferian outlook in reality that makes our world uglier, higher friction, and more coarse.
People having cool ideas and sharing them make our world brighter.
Super well said - right. “Try again with quick feedback” vs “try again with significant feedback” vs “try again, but only a subset of the original task” vs “let’s have someone else do this”
Cosign on the AirPods Max. I'll easily use them daily for 4 years, and then someone else will enjoy the hell out of them.
It makes music even more beautiful.
The noise cancelling is absurdly good.
Mic quality is great.
They're incredibly comfortable for long work sessions.
The little buttons and dial are a joy to click and press.
The case everyone made fun of? It's awesome. By being forced to put them away, it always shuts off when not in use. The mic doesn't drain like every Bose I've ever ad. Fantastic product.