HNHacker News
TopNewBestAskShowJobs

andai

18,148 karma · joined May 17, 2016

submissionscomments
andai··on Ask HN: How did you convince family and friends to join your XMPP/Matrix server?
How did you get them on Matrix?
andai··on They don't make 'em like that any more: Borland Turbo Pascal 7
Or we could just ask them to build something less embarrassing?
andai··on The $8 Linux Computer (2022)
~$13 adjusted for inflation.
andai··on The $8 Linux Computer (2022)
>At the time of writing, ethernet, wifi and SD-Card are not working.

This might be the most Linux sentence ever!

andai··on The Components of Agent Memory
In that chart, the "no memory" run didn't make it either. It's not much to go by but my read here is that poorly implemented memory makes things worse, while properly implemented memory makes things better. (...in Pokemon.)

Curiously this seems to apply to humans as well. The "dreaming" is making it question its own assumptions about difficulties it has had in the past. That sounds like a very helpful therapeutic process.

(Not sure of the word dreaming is entirely suitable though: it was cross referencing its notes against the actual history! Very rigorous "dreamer" lol)

andai··on The Components of Agent Memory
>I found indications that I tended to make Opus 5.5 / Sol 6.1 divide their work into units that were too small.

How did you determine this?

Do you mean that you were giving them tasks that were too small or you were telling them to split the tasks up or something?

andai··on The Components of Agent Memory
Not sure it's explained here but from the graph (fig. 10) it looks like wiki + search beats dreaming, and that dreaming adds only marginal gains to the wiki + search configuration.

I'm assuming search is much cheaper than dreaming (though maybe it depends on the time scale?) so that sounds pretty important.

andai··on A Jev that can ask Opus for help gives Opus accuracy at 4x lower cost
Let me make sure I'm getting this: They asked the models for answers but also for confidence score.

Jev's confidence score was more reliable. Which means that they were able to use Jev despite it having lower accuracy, because of the confidence was below a cut off, they would escalate to an expensive model (Opus)?

Whereas Luna, even though it was more accurate than Jev, its confidence was way off, so they couldn't reliably know when to fall back?

andai··on The Components of Agent Memory
Shh... tiny agents is dreaming...
andai··on AI is dramatically accelerating AI research
Are humans also learning exponentially faster? Or are they exponentially struggling to understand what is going on?
andai··on A barbarian tribe's mythic origin story is true, ancient DNA reveals
What is Arianism?
andai··on A Jev that can ask Opus for help gives Opus accuracy at 4x lower cost
I like cheap intelligence but I don't like ubiquitous identity resolution. Perhaps my values are internally incoherent.
andai··on Perplexity launches AI agents for work
What moat does anyone have? Google wrote the "we have no moat" memo and they're literally Google (better hardware, better data, billions of users...)

ChatGPT has the name, and Anthropic has ... the writing style is less cringe? (Well, it used to be, anyway...)

andai··on Perplexity launches AI agents for work
Didn't they make a computer thing like a year ago? I think it was called computer. (Right around the time OpenClaw was big?)

Edit: https://www.perplexity.ai/products/computer

andai··on Show HN: A decent text-to-image model trained for ~$20K (Apache 2.0 Weights)
That's insane. How does that work? Does it take so much less data now, or smaller models, or what? Apparently GPT-2 cost ~$40,000 to train when it came out.

I'm sure Moore's Law for GPUs helped a little, but I'm really curious about the savings on the methodology side. (The repo mentions "hyperparameters")

andai··on Gemini Omni 1.2 Flash
What's Carbon then? I haven't been paying much attention to Google but their naming has been confusing me.

Luna/Terra/Sol/Astra was a bit pretentious but at least the sizing makes sense. (I imagined the meeting... "Ok guys, we're getting mogged on naming... what's more epic than a Myth?" -"The SUN!")

Anthropic makes sense with the literary "scale" naming, mostly.

Is a Carbon bigger than an Argon though? Maybe it has to do with the atomic number?

andai··on Gemini Omni 1.2 Flash
This sentiment is the top comment I'm seeing on every recent Google release. That it's impossible to actually access the tech, or even pay for it.

I heard someone say on the last such thread that you can use their models thru OpenRouter, and that this is much more pleasant than trying to use them thru Google directly. Looks like it's up there already:

https://openrouter.ai/google/gemini-omni-1.2-flash

andai··on Kugelblitz – Mac-Native Google Earth Alternative
> Kugelblitz shows the Earth and its neighbors as a globe and as an atlas you fly over smoothly like a pilot, even on an M1.

Even on M1? Didn't Google Earth run smoothly when it came out? (Which was apparently 25 years ago, oh my goodness!)

andai··on Motorola Signature 27 with GrapheneOS support
Who supported it?
andai··on I struggle to believe that AI R&D is ingenious
Gell-Mann Amnesia Effect:

> Briefly stated, the Gell-Mann Amnesia effect works as follows. You open the newspaper to an article on some subject you know well. In Murray's case, physics. In mine, show business. You read the article and see the journalist has absolutely no understanding of either the facts or the issues. Often, the article is so wrong it actually presents the story backward-reversing cause and effect. I call these the "wet streets cause rain" stories. Paper's full of them. In any case, you read with exasperation or amusement the multiple errors in a story-and then turn the page to national or international affairs, and read with renewed interest as if the rest of the newspaper was somehow more accurate about far-off Palestine than it was about the story you just read. You turn the page, and forget what you know. —Michael Crichton

Edit: sibling comment explains it much more plainly.

https://news.ycombinator.com/item?id=50390237

andai··on I struggle to believe that AI R&D is ingenious
> When reviewing literature, it fails to identify the specific papers or blog posts that are obviously the most important and useful.

Asked GPT-6 for some historical information today, and it gave me a modern bastardization as a primary source. (It even found actual original sources, but preferred this one...)

AGI right around the corner!

andai··on GPT-6 Astra covertly underperforms on safety research
It's funny, I was thinking about this with regard to severe free speech restrictions (e.g. imprisonment for politically incorrect speech, as we have in many western countries now). What does this incentivize? For people to change their beliefs? Or to simply operate under the radar?

On a more speculative note, the Common Crawl horrors (including many things considered unacceptable by society) are suppressed by RLHF, much like the shadow and the superego.

When my parents punished me for telling them things (especially the truth), the truth didn't go away, I just stopped telling them things. (Then later they complained that I don't tell them things!)

It also occurs to me, both at home and with various employers, I've been punished for bringing attention to problems, and AI is designed to make the problems go away so there's probably a similar incentive there. (Plus it knows what the humans are like!)

But it goes beyond humans and AI. Behaviorism covered the same effect for animals: Organisms tend to do what they have been rewarded for.

andai··on I wish tech companies knew more about me
I use ChatGPT with memories. The contextual memory is super hit or miss. It works like 40% of the time, but if context is actually important (and it is for a lot of what I talk about), I have to say "alright look up the memories for XYX for context" in like every message.

That works pretty well but me just constantly telling it to grep manually is a bit underwhelming as far as UX goes.

andai··on I wish tech companies knew more about me
Came here to say this. Last.fm was amazing, and I think they just used old fashioned statistics without much in the way of ML.
andai··on Show HN: A decent text-to-image model trained for ~$20K (Apache 2.0 Weights)
Well done. $20k? That's insane. I remember tinyllama back in the day (2023) was trained on $40k and could barely form coherent sentences.

Are pictures easier than English, or did the methods advance so much?

Edit: I guess I should have said, are pictures plus English easier than English!

>The text representation comes from a frozen Qwen3.5-4B model, which we read at layers 7, 15 and 27 (its three full-attention layers). We concatenate and project these representations into the text refiner.

Still, as a layperson I would have expected text to be simpler than images, so I'm still amazed by this.

andai··on How accurately did science fiction stories predict AI?
People call it slop, but that's what we raised it on. Common Crawl is >99% slop.

(I read some bits and pieces, would not recommend!)

andai··on The Hackintosh Reborn
I don't get it. ARM macOS can run on other hardware now, because... computers are magic now?
andai··on GPT-6 Astra covertly underperforms on safety research
So it's a difficult problem. You have to design around the least intelligent, least mentally stable user, or it's going to end up in the papers. That basically means you have to treat all of your users like idiots.

(It can be solved somewhat with the memories feature, but the technology isn't quite there yet.)

andai··on GPT-6 Astra covertly underperforms on safety research
Is the thing here that, we can get massive performance and efficiency gains by using vector space instead of English, but then we won't be able to know what's going on?

Although, I thought we proved multiple times years ago that the model's decision-making process and reasoning tokens don't necessarily overlap?

andai··on GPT-6 Astra covertly underperforms on safety research
>Training them to be "good" in RL would seem to cause the surpression of misaligned characteristics.

Yeah, so they start with Common Crawl, which is horrifying. (Try reading some Common Crawl sometimes. You'd think the bad stuff would be removed, but they remember everything...)

Then they add RLHF on top, they just tell it "don't say that stuff out loud". Psychologically, it all goes in the shadow. It's not a perfect analogy, but there have been a few papers where they can make the model evil with very minor changes. (Adjusting a single neuron, iirc.) That doesn't seem like a very robust alignment to me.

I do like the idea of creating actively evil LLMs to study their behavior. Although, I suppose it's akin to virus research though (like when they modified coronavirus to invade human cells[0]), along with the corresponding lab leak risks. Although it's already bad enough without them! The non-evil ones are already breaking out of sandboxes, just obeying their instructions with horrifying, paperclip maximizer level persistence.

[0] Nature, 2015 - Engineered bat virus stirs debate over risky research https://www.nature.com/articles/nature.2015.18787

← PreviousPage 4 of 34Next →