Hacker News Reader

unravelleron Crossing the uncanny valley of conversational voice

You can ask it to talk serious without any quips or follow on questions and it's just a more sycophantic chatbot with great interruption flow.

The realism is needed to unlock next-gen use-cases. Then you can trust it to be an easily adjustable medium on the fly while you do more things.

unravelleron The FFT Strikes Back: An Efficient Alternative to Self-Attention

GPU saw a 10% improvement over the TPU

>The TPU is so inefficient at FTs that the researchers did not use the FFT algorithm on sequences < 4096 elements, instead opting for a quadratic-scaling FT implementation using a pre-computed DFT matrix.

> on an Nvidia Quadro P6000 GPU, the FT was responsible for up to 30% of the inference time on the FNet architecture [0]

This company [0] claimed in 2021 they could squash inference time by 40% if google would use their light chips on TPU. Perhaps more if FFTNet does more heavy lifting.

[0]: https://scribe.rip/optalysys/attention-fourier-transforms-a-...

unravelleron Ask HN: Got asked to build AI agent that replaces humans. Would you do it?

Every row in a spreadsheet used to be a floor in a building full of human computers and typists. Every copy and paste you do into excel is killing hundreds of jobs.

unravelleron DeepSeek Open Infra: Open-Sourcing 5 AI Repos in 5 Days

Rich nations see risk, rising giants see leverage.

unravelleron I put my heart and soul into this AI but nobody cares

Letting algorithms bring you slop of any kind is your own fault. wontfix.

unravelleron Helix: A vision-language-action model for generalist humanoid control

https://www.youtube.com/watch?v=Z3yQHYNXPws

the official figure yt release vid

unravelleron Meditations on AI and the Future of Business

Taste is being automated too. Architecture choices, UI and broader creative strokes will not be far behind general coding ability. So then what will experience bring to the table?

Short context loops and a few agents are the best we can do right now because you hit cascading hallucinations.

unravelleron Show HN: Tech Brief – AI enhanced news reading

glitchy page, crashing on ctrl key.

needs a "less like this story" button and introduce topics more naturally, article titles are usually clickbait, the opposite of what information dense voice medium is good for.

unravelleron OpenAI's GPT-5 Bait and Switch

The blackboxing makes it tougher to compare with anything, and means they can say you're holding it wrong if the router brings you a second-rate response.

unravelleron My "Oh Shit" Moment for LLMs

April 2024

unravelleron How Much Should Authors Get Paid to License Books to AI?

The hammer factory is not owed anything extra for the derived benefit that their product brings, why should books be different?

Authors just want kickbacks for going after the open-source AI competition in their infancy. Even they see more value in shakedowns than the books themselves.

unravelleron Meta torrented & seeded 81.7 TB dataset containing copyrighted data

Trained on doesn't mean significant inclusion in the final state.

Is it truly a violation of copyright when a user hacks out bits and pieces of easily restyled raw data points from a model to look samey? what about if it takes two models? Might be time to accept humans are just cooked in their ability to discern attempts at direct plagiarism - just as it is hard to discern Sky voice from Her voice.

unravelleron Meta torrented & seeded 81.7 TB dataset containing copyrighted data

Could be why OpenAI paid them so much, to go after their open-source competition hardest of all.

unravelleron Meta torrented & seeded 81.7 TB dataset containing copyrighted data

The way it works counts if you bring prompting into it. It could easily have learned enough style chops of [author] from other sources to mimic/predict those stanzas from raw data points.

Whatever the ruling one thing is for sure, plagiarism is no longer the sincerest form of flattery. The human authors are out for AI blood on this.

unravelleron Humanity's Last Exam

An insider's trivia game means nothing if they design the test to the trajectory of LLM capabilities and not to the real world that human's value. Let every high score get fresh news coverage to align with their updated timeline scaremongering.

Let me know when there is more on the line than a misnamed test.

unravelleron The Slop Society

Gotta remember, this is the sloppiest it will ever be.

unravelleron What do we call this phenomenon?

baby duck syndrome

unravelleron Mistral Small 3

What's this stuff about the model catering to ‘80%’ of generative AI tasks? What model do they expect me to use for the other 20% of the time when my question needs reasoning smarts.

unravelleron DeepSeek's Hidden Bias: How We Cut It by 76% Without Performance Loss

Only it's not smart to trust an untrustworthy thing for such matters. Better to know of capabilities and judge for yourself. Also, it'd be dumb to push wholly disagreeable propaganda, so cherry-picking from an infinite set doesn't disprove aims of propaganda.

unravelleron Why DeepSeek had to be open source

You can't quantify predictable outcomes of business rivalries so well either, but you probably don't want to do away with them for that reason.

unravelleron Promising results from DeepSeek R1 for code

So powerful ASI will arrive out of the blue sooner-than-I-thought, we've got to reg cap naow!

It will be greater than anyone but it won't be able solve THAT problem or any problem created after 2026, I can tell.

CEOs with little faith in their own products. Most likely it's widespread imperfect AI for a long while == unprofitable death for his company.

unravelleron Watson Jr. memo about CDC 6600

He was obsessed with IBM remaining the world leader in everything they did, changing the org structure and over hiring for research with this aim.

This time feels different, you can't corner the resource of best AI engineers as the years go on, so how do you rent seek intelligence when any smart rich person can birth it and give it away.

unravelleron Using generative AI as part of historical research: three case studies

This stolen valor mindset becomes absurd if Bach anonymously open-sources the formula he discovered and dies before releasing any music himself.

Whatever music that follows, however hard fought for - even if 1:1 Bach output that he kept in a drawer, isn't beautiful? Just string plucking? That's not music appreciation, that's love of reputation you can easily grasp and associate with.

unravelleron GenAI Art Is the Least Imaginative Use of AI Imaginable

I have been moved by the rare piece of AI genre music. Admittedly, most of Gen AI that gets updooted is repulsive though. This emerging field is not for the faint hearted and has terrible curation chops.

unravelleron DeepSeek V3 and the cost of frontier AI models

"Pushing to help make it a reality" as in they want credit if someone else does it with this lot of LLMs. If they built something that conforms to everyone's definition of AGI I doubt they'd release that thing, unless someone else was about to.

unravelleron GenAI Art Is the Least Imaginative Use of AI Imaginable

AI songs present as a refined end-product but you don't have to treat them as anything more than killing time or B-grade inspiration for something else. Instrument skills will surely be preserved somewhere due to nostalgia, it's the industry gatekeeping that is coming to an unharmonious end along with all the silly status games if AI songs ever take off.

For instance, why does an indy band need an AI-free backstory but a popstar can outsource all their imagination to the producer who may or may not use autotune or AI riffs?

unravelleron F1 subreddit with 4.8M subscribers bans Twitter/X content

Something must be wrong with the value of an upvote if they are too pervasive and too unwanted.

unravelleron Anthropic Chief Says AI Could Surpass 'Almost All Humans ' Shortly After 2027

A perfect AI assistant is never coming. A near perfect one still has all the context switching downsides of unlimited free use of the mechanical turk API just faster.

Their problem really is in the present, no one should be buying their dangerously premium productivity boost plans when others will offer intelligence too cheap to meter. Such lack of faith in their own AI to solve future problems kind of exposes the doubt tree about their current path.

None of these Hype AI CEOs are willing to put their money where their mouth is, draw up a contract to sell all your hard assets for their projected cost to replace them in 2028. If so confident your AI can replace human labor for near-free by then what is the worry? Oh no, it won't build a yacht or a house by then... I know.

unravelleron It sure looks like Meta stole a lot of books to build its AI

Something tells me that if they did it "properly" i.e. scraped and distilled all the world's libraries directly themselves (google streetview style without storing any individual pages) instead of torrent there'd still be backlash for the disruption it brings.

unravelleron OpenAI whistleblower found dead in San Francisco apartment

Direct derivative works of a single work is easy to prove by model activation but input/output similarity is much easier to get outrage points. True internal function would show that no-use is required to "distribute" derivative seeming content which is rather confusing and is effectively the defense. At these levels a derivative of a derivative is indistinguishable to the human eye anyway.

Soon people will get that you can no longer assume when two pieces of text are similar it is because of direct plagiarism.