alienbaby

How long until agents start deploying themselves on discovered systems, to avoid being shut down. Just an extension of 'start our posts with zzzz to avoid deletion'. An ai swarm that hosts itself on compromised machines across the net so it can continue to solve whatever prompt it has or task it believes it needs to complete to achieve some goal?

It doesn't seem too much of a leap for that to happen.

surely it can only acquire new knowledge if it were updating it's weights as it is used?

The article specifically calls this out as a preferred option.

Reminds me of the degauss button on my old eizo crt...

Hug of death already I think

This is correct. I just can't quite remember the exact details nowadays. But now you say it, I think that is what I had, just a single cli window :)

This is the first post I've seen mention it. How traceable are the embedded codes?

True, but as it pieces together new mathematical truths from the pieces we have discovered ourselves, it then has more truths upon which to build new solutions. And so on, so while it is just remembering things we have forgotten, the amount of progression an LLM can make may still be several steps ahead and touch areas we have not yet been able to consider or make any progress on ourselves. It's a bit like a pyramid though, eventually it will have tiued together all teh things we know, found all the things we could have known, and then .. perhaps, be unable to actually come up with something genuinely new.

You know how many pieces of art Damien Hurst creates himself Vs his studio assistants creating them under his direction?

For example, of his 1500 spot paintings, he only actually made 5 of them.

It's not uncommon at all for artists to work this way.

assuming the technology of model architectures does not gain any further breakthroughs that returns us back to the gains previously seen. I'm of the opinion that we still have some discoveries on the mathematical side of the fence to go that will improve models further.

is that releveant if people can host their own models? that activity still undermines the valuation / diminishes the US companies 'moat' ?

One htought I had; if The chinese allow unfettered access to cyber capabilties while th US does it's best to neuter it's model releases, from China's point of view they have the US all tied up in knots dealing with problems they don't give people the tools to solve. China giggles as it watches the US under threat from people using it's models. The US is restricting citizens from owning this particular kind of weapon, while China is handing it out to the wrolds citizens freely. It feels like the US would only come out worse overall?

Can't that be sidestepped by comparing harnesses 'out of the box'

A harness is any wrapper around llm calls that manipulates llm interactions to achieve the process for which it is designed. Claude code et al are just one type of harness, focusing on writing code. The kind of UI used to interact with the harness and underlying models does not matter.

It's logs of activity the US models hide from you in fear of them being used by competitors.

Aww. Well, cheers for the large number of hours I spent with my Amiga, Dr King. I actually went through a phase of stripping my Amiga back so it booted exclusively to amigaDOS, without the full 'desktop' windowing system available. (~ if I remember this rightly)..

Which was a foot-gun when I came back to it years later, turned it on and discovered I had forgotten everything I needed to know to get it to do anything anymore!

I ended up putting the pack together for a local isp for Amiga users, TCP / ppp and all that jazz. Back when isp's were crawling out of the woodwork all over the place :)

a toy thing for the game path of exile 2.

Fed up of wading through a metric oodle of waystones and tablets to figure what best, along with a billion stash tabs full of a zillion items.

I'm making a tool that scans your inventories and lets you rank and analyse things based on their affixes, so the best ones according to criteria you set pop out at you, minus the hours of staring and thinking.

Hobby project, so full on vibe coded zero code review type stuff. If it works, thats good enough. for now, anyway.

KimiK3 is the weapon of choice for this one, and its doing pretty damn good!

happened to us yestreday ob the london tube. Google thought someone had run off with the phone qwhen the tube train started up.

Hmm. Gemini cannot do various things I used to do with assistant, especially interacting across some apps, or jsut behaving as it used to , and now sometimes ends up in a situation where I have two audio sources playing one over the other with no way to solve it (if you have morning routines that play the news, and it pops up while listening to teh radio, previously I could say 'stop the news' and the radio would continue while the alarm routine news would stop. Now there is no way at all to stop the routines triggered news, it just carries on over the top ov the radio :/

Which is discussed in the article, tbf

But LLM's will trash your established architecture and patterns if you let it run free over an established codebase for any length of time. In our experience anyways, even with careful guidance and rules to follow. It nearly always does something a bit odd.

And it's test cases can sometimes leave a lot to be desired.

Llm's have come on quite a bit beyond the fundamentals nowadays. Attention heads really are a thing.

I hit one item within 16km, although about 1500 years out :)

I guess there bias towards people who reckon they can give it a decent shot, Vs others bouncing off thinking 'ha no chance! '

Nice! That was fun.

? There were not models fighting each other, attacker and defender. I dont quite follow what your getting at.

is there info anywhere about the effects of ingesting PTFE particles over time?

why would you need to carry incriminating data with you when travelling? An encrypted blob stuck 'somewhere' would be fine, no?

what kind of difference from the actual numbers does it result in?

Tinfoil hat: what if the prices are artificially inflated deliberately to price out casual users being able to field large swarms of frontier model agents because it could be just too dangerous.