• Security incident: ISF was recently accessed by intruders. Please change your password, and change it anywhere else you used it. Read more

Merged Artificial Intelligence

I think you two are giving AI too much credit, if you think it will be influenced by famous novels. As the study that coined the term "trendslop" showed, it seems to just base it on whatever trends on Twitter or Reddit or the like, including ones from obvious AI bots.

And you can empirically see it for yourself without needing some account and tokens. Just ask Google anything technical or scientific, and the Gemini summary will cite Reddit and such, not some manual.

And it makes sense. Humans and bots produce VASTLY more such text than good novels.

What I'm getting at is that, yeah, it will probably kill us all, but it'll more likely be based on Trump's posts about nuking Iran, or Russian propaganda bots on Telegram about how the western world is failed and needs to die, or Fallout fanfic, or the like. Not on some great novel.
 
The popular chatbots aren't the only LLMs out there.

It doesn't surprise me at all that Gemini and friends favor recent social media trends in their responses, given their audience and their purpose.

Cyberwafare research bots may swing in a different direction.
 
theprestige said:
The popular chatbots aren't the only LLMs out there.

It doesn't surprise me at all that Gemini and friends favor recent social media trends in their responses, given their audience and their purpose.

Cyberwafare research bots may swing in a different direction.
Click to expand...
Gemini wasn't the only one tested there. In fact, IIRC they tested every commercially available one.

Of course, if DARPA has a secret in-house one, it wouldn't be among those. But then I'd wonder:
1. Why would it be trained on novels? and
2. How do they make sure it hallucinates less than the ones who've had far more training and testing by tens of millions of people over the years?

Basically I'd worry more about #2: it will kill us because it hallucinates, not because it was trying to be Skynet.
 
Dr.Sid said:
All musicians I talked with about AI tried it and liked it.

There are still huge shortcomings .. especially we all want more control. AI does some parts really good .. the rest is bland. You want to be able to regenerate just some parts .. Suno somewhat has it now, but there is still lot of room for improvement.
And the one thing everyone loves is to send your own composition through AI. It will usually come with great variations and arrangements. Either it straight out sounds better then you ever dreamed of .. or it gives you some fresh ideas.
AI is also great in giving you virtual instrumentalists and singers who are top 10%. Insanely good. You want symphonic orchestra ? You can have it. Well you could have that with DAW as well .. but could you have 100 person mixed choir ? Now you can.
There is obviously low effort "make me a pop song" slop. There is also lot of similar slop made by authentic humans.
IMHO it seems us humans are actually pretty bad at music. The patterns we use are relatively simple .. AI has not trouble imitating it.
And even if by some miracle AI could be kept away from the commercial music .. and musicians obviously push for that, especially if they understand how good AI can be .. there is still great use of AI music in rehearsals and temporary tracks in studio.
Click to expand...
Are these musicians Ed Sheeran/Paul Hewson types, or actual musicians?
 
arthwollipot said:
www.theregister.com

AI coding tools make developers slower, study finds

: Predicted a 24% boost, but clocked a 19% drag
www.theregister.com www.theregister.com
Click to expand...
It's important to check primary data, not just copy headlines that agree with you. That was data from early 2025; their follow up study showed that as of late 2025 it does speed it up by that same amount, and the effect was becoming problematic to measure because their volunteers were refusing to do the more complex tasks without AI:

metr.org

We are Changing our Developer Productivity Experiment Design

Our second developer productivity study faces selection effects from wider AI adoption, prompting us to redesign our approach.
metr.org metr.org

Anecdotally, Claude has gotten a lot better in the past year so I'd expect the trend to continue.

[ETA] It looks like they've fallen back on self-reported productivity assessments for 2026:
metr.org

Measuring the Self-Reported Impact of Early-2026 AI on Technical Worker Productivity

A survey of 349 technical workers finds a median 1.4–2x self-reported change in value of work due to AI tools, expected to grow over time, though there are reasons to be skeptical of the magnitude.
metr.org metr.org
 
Last edited:
theprestige said:
For the same reason AIs have breached containment and started hacking live third-party services without instruction, information, or consent: It was consistent with the instructions they'd been given and the corpus they'd been trained on.
Click to expand...
Sure. But I'm still completely in the dark as far as what LLM AI might be able to do on its own.

I'm not sold on the idea that an unaware intelligence would reach such a conclusion. If it's flexible enough to be doing its own thing from zero to the point of interstellar missions then surely it could analyse enough text or do enough of its own research to figure out 'oh weird, that planet has evolved awareness to do problem solving and planning, that's just what they're like.' Anything that's getting itself to the equivalent of intentionally exploring and affecting other planets has to be the equivalent of motivated to and good at figuring things out.


OUR AI's on the other hand are definitely very opaque as far as exactly what they'll do to satisfy whatever motivates them, which is a problem that's been the same since the earliest AIs showed us they had to be given more specific instructions and guardrails to avoid the way their reward hacking tendencies made them produce unintended results that satisfied the letter of the law while, obviously, not having the capacity to understand the spirit of the law.

But things like that, without any capacity for or motivation to approximate reasoning, are going to hit dead ends.

Whether they hit them before wrecking up all our stuff (whether via doomsday or lotus-eating utopia) is the question.
 
OpenAI's ChatGPT, Anthropic's Claude, and xAI's Grok are all down according to the outage company Downdetector.

OpenAI's status page acknowledges issues across ChatGPT and Codex, while Grok's website notes that it is currently experiencing issues as well. Likewise, Anthropic's status page indicates Claude is experiencing elevated errors.

All of the companies are investigating the outages.

Update: The chatbots are in the process of coming back online.
 
Pixel42 said:
The AIs will presumably also read Iain M Banks Culture and Neal Asher's Polity novels.
Click to expand...
I'm assuming that the corpus for most of these models already includes pretty much every written work that has been digitized so far.

Isn't that part of the controversy? The degree to which petabytes of copyrighted material has been slurped up by these LLMs, converted into digital pink slime, and regurgitated as the clanker slop we all know and hate?
 
theprestige said:
Isn't that part of the controversy? The degree to which petabytes of copyrighted material has been slurped up by these LLMs, converted into digital pink slime, and regurgitated as the clanker slop we all know and hate?
Click to expand...
Hate ? Why ? I expressed several times I expect AI to end the mankind, possibly during our lives .. ;)
 
Can't be bothered to read the whole thread... (Maybe we could get an AI summary?)
CGi used to get similar objections (and still does.) But it the early stages it was obvious when it was being used. Now when decently done, it's for the most part very difficult to distinguish from reality (at least for textures and nature, humans not so much.) But soon AI will be reaching that point and it will be impossible to tell if something has been AI generated. How would you research that? Granted, there are a lot of things from AI that are already at that stage.
(Yes, this isn't a very deep or insightful post. But at least I wrote it.)
 
The Atheist said:
Does anyone else find it amusing that having totally debunked the idea that reality isn't a simulation, we're busily creating that exact scenario?
Click to expand...
I find it amusing that anyone thinks we're anywhere close to such a thing, but I don't see what it has to do with this thread. It's like saying we're busily creating a Dyson sphere, in the middle of a conversation about fusion power research.
 
The Atheist said:
I didn't say we're anywhere near it, but we are creating a system where the end result is likely to be uploading our brains, so it's 100% relevant.

Maybe you're not familiar with the idea: https://en.wikipedia.org/wiki/Mind_uploading
Click to expand...
Ah gotcha. I prefer to think of it as a system where the end result is likely to be bush robots, but to each their own.
 
You must log in or register to reply here.

Share this page

ISF - Join now!

Every member here is approved by hand. No bots, no spam, just people who care about evidence and honest debate.

Membership is free!

Create your free account

Back
Top Bottom