HNHacker News
TopNewBestAskShowJobs

uh_uh

763 karma · joined December 7, 2020

submissionscomments
uh_uh··on 2023: Focusing on a single product pays off
If your SaaS can be killed by a single customer leaving, then you don't have a SaaS; you have a boss for whom you are developing bespoke software.
uh_uh··on Is my plane a 737 MAX?
I don't think it's worth celebrating an emergency door being blown off a plane that already killed 300+ people years prior. This is not how the system should work.
uh_uh··on 2023: Focusing on a single product pays off
In a typical employment scenario, your line manager can cut off 100% of your revenue.

In an ideal SaaS scenario, your customer can cut off a tiny percentage of your revenue.

uh_uh··on 2023: Focusing on a single product pays off
The difference is that some bosses are more easy to say no to than others.
uh_uh··on Three senior researchers have resigned from OpenAI
> I agree AI is useful, but not to that extent to what it is valued on the market.

I agree, it's greatly undervalued!

uh_uh··on A coder considers the waning days of the craft
If LLMS are not on the path to general intelligence, what is then?
uh_uh··on Sam Bankman-Fried Convicted
And when AI does the same thing, some people argue that it's proof for its lack of intelligence.
uh_uh··on Ask HN: SaaS Founders, What 3 advice would you give your younger selves?
How do you sell when you have nothing but words?
uh_uh··on John Carmack and Rich Sutton partner to accelerate development of AGI
I mean, an ant has signs of life yet you wouldn't call it sentient.
uh_uh··on The Return of a Small Universe?
That matter from the other universes seem to pretty well line up with galaxies in our own universe which makes this theory a bit sus. I'd expect at least some of them to be scattered at random places, whose gravitational lensing effect we should be able to observe.
uh_uh··on Octopuses, crabs and lobsters recognised as sentient beings under UK law (2021)
Sometimes I wonder how future generations will judge our meat industry practices once they figure out artificial meat production and also have a more accurate picture of consciousness/sentience and hence the suffering we cause. It's quite plausible that their judgement will be just as harsh, if not hasher, as how present day humans look at "savages" of the past.
uh_uh··on Octopuses, crabs and lobsters recognised as sentient beings under UK law (2021)
I'd be surprised if there is any mammal that doesn't.
uh_uh··on Why the original transformer figure is wrong, and some other tidbits about LLMs
I'm still not clear on the second question. If lalaithion's original statement "the Q and K matrices aren't structurally distinct" is true, then once the neural network is trained, how can we look at the two matrices and confidently say that one is the query matrix instead of it being the key matrix (or vice versa)? To put it another way: is the distinction between query and key roles "real" or is it just an analogy for humans?
uh_uh··on Sam Altman goes before US Congress to propose licenses for building AI
I'm not suggesting that they have any. I was reacting to srslack above making _technical_ claims why LLMs can't be "generalized and adaptable intelligence" which is not shared by said technical experts.
uh_uh··on Sam Altman goes before US Congress to propose licenses for building AI
2 reasons:

1. He's too busy building the next generation of tech that HN commenters will be arguing about in a couple months' time.

2. I think Sam Altman (who is addressing the committee) and Ilya are pretty much on the same page on what LLMs do.

uh_uh··on Sam Altman goes before US Congress to propose licenses for building AI
srslack above was making technical claims why LLMs can't be "generalized and adaptable intelligence". To make such statements, it surely helps if you are a technical expert at building LLMs.
uh_uh··on Sam Altman goes before US Congress to propose licenses for building AI
Both of these statements can be true:

1. ChatGPT knows the algorithm for adding two numbers of arbitrary magnitude.

2. It often fails to use the algorithm in point 1 and hallucinates the result.

Knowing something doesn't mean it will get it right all the time. Rather, an LLM is almost guaranteed to mess up some of the time due to the probabilistic nature of its sampling. But this alone doesn't prove that it only brute-forced task X.

uh_uh··on Sam Altman goes before US Congress to propose licenses for building AI
I just don't get how the average HN commenter thinks (and gets upvoted) that they know better than e.g. Ilya Sutskever who actually, you know, built the system. I keep reading this "it just predicts words, duh" rhetoric on HN which is not at all believed by people like Ilya or Hinton. Could it be that HN commenters know better than these people?
uh_uh··on We aren't close to creating a rapidly self-improving AI
Previous approaches like LSTM struggled learning long-term dependencies. The transformer improved on this greatly.
uh_uh··on Is there a canonical source for “the argument for AGI ruin” somewhere
I think you're talking about ASI (artificial superintelligence). AGI just means that it's a general problem solver as opposed to a narrow AI. It doesn't have to be superhuman.
uh_uh··on Show HN: YakGPT – A locally running, hands-free ChatGPT UI
I did, just not in the JavaScript ecosystem.
uh_uh··on My kids and I just played D&D with ChatGPT4 as the DM
I'm waiting for the moment when an LLM will prove a new math theorem or something and people will go: "Well, you know, it's not really that impressive considering it used X and Y previous results produced by humans. The model's contribution is tiny. This proof was kind-of low-hanging, a lucky PhD student would have stumbled into it eventually." It's coming.
uh_uh··on ChatGPT Plugins
Yes, from what I understand, these follow a similar model as Shopify apps.
uh_uh··on Build full “product skills” and you'll probably be fine
> You're assuming that the AI is even generating anything that will make sense to a human.

Why wouldn't it? It's trained on code generated by humans and already generates code that is more readable than the output of many humans me included.

uh_uh··on Alpaca: A strong open-source instruction-following model
Has this actually been tried? It would be trivial to generate a very large dataset of simple additions and see if an LLM gets it. Maybe it wouldn't serve an immediate practical purpose but I'm curious.
uh_uh··on Visual ChatGPT
Maybe another way to think of it is that the error correction part of image generation models is offloaded to the human visual cortex which is a very old evolutionary construct and thus had time to become very resilient? In case of text generation, maybe the error tolerance of the human brain is less developed as human-level language is a newer evolutionary invention.

It'd be interesting if the parameter/complexity requirements are actually similar once you examine the system as a whole, meaning machine _and_ human brain.

uh_uh··on 7 Weeks AI Alignment Curriculum
Part of the issue is that taking AI alignment seriously does require some level of intellectual humility — a quality that the HN comment section famously lacks.
uh_uh··on Theory of Mind May Have Spontaneously Emerged in Large Language Models
What else would it be? Are you suggesting some non-physical process?
uh_uh··on Show HN: StackOverflow.gg – AI-generated answers to every coding question
Cache a couple of them. Let people vote on the best.
uh_uh··on OpenAI-to-SQLite
OpenSearch looks like the best so far, all my requirements combined!
← PreviousPage 5 of 9Next →