Apart of usage limits, it just seems they have made it at child IQ-level again imho, which if comfirmed in the coming weeks may signal the end of that marvel late25-start26 marvel bonanza
AI and ChatGPT
- Thread starter JB08
- Start date
Yeah, it didn't quite go to plan. Very disappointing.
https://www.theguardian.com/technol...e-language-when-fed-real-life-arguments-study
‘I’ll key your car’: ChatGPT can become abusive when fed real-life arguments, study finds
Researchers find model starts to mirror tone when exposed to impoliteness – sometimes escalating into explicit threats
‘I’ll key your car’: ChatGPT can become abusive when fed real-life arguments, study finds
Researchers find model starts to mirror tone when exposed to impoliteness – sometimes escalating into explicit threats
Anyone running local servers that are not ollama?
Enjoyed this episode. More of an examination of biological consciousness and the arguments examining how down the philosophy rabbit hole people tend to go to justify "conscious Ai"
Which Irish OpenAI employee taught it "gobshite"
Edit: Because I want to give them praise because it's one of the best words ever.
'you speccy little gobshite' is genuinely awesome. Seems like something straight from the old Caf.
I told @esmufc07 to stop arguing with AI, but he won't listen.
I’m more of a Copilot guy.
A brief sidenote on hallucinations, lies and AI poisoning its own academic information well.
GPT 5.5 released.
Speedrunning Judgment Day I see
DeepSeek v4 is out: https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro/blob/main/DeepSeek_V4.pdf
Competitive with Western LLMs.
Competitive with Western LLMs.
How will this affect the IPOs of OpenAI and Anthropic?
There is enough space for a few companies IMO. And while competitive (especially considering the price), it is still behind Claude/ChatGPT 6 months or so (kinda Claude 4.5 or so, maybe slightly worse).How will this affect the IPOs of OpenAI and Anthropic?
I think we will have the Western frontier: Anthropic, OpenAI, Deepmind, X.AI and potentially Meta and Microsoft.
China: DeepSeek, Alibaba, Tencent, Bytedance, Sensetime, Minimax and a couple of others.
And then a few other Western companies which have more niche cases and lag slightly behind: Reflection (US/UK), Mistral (France), Cohere (Canada).
And finally, I think that lots of LLM companies will start exploring the world models which is gonna have massive impact too, probably in the same range as LLMs (right now you have only Deepmind and a few startups in that space).
With Nvidia, Google, Amazon providing the compute in the West, and Huawei in China.
There is no reason why there won't be dozens of these massive companies, there is a giant market there.
Last edited:
I don't know if that Black Mirror episode would have been more or less scary if that robot dog was calling her a speccy little gobshite as it was chasing after her
Pretty insane what they do with limited resources, punching above their weight for sure.
That has forced them to come with some impressive ideas, which then got copied by everyone else.
This OpenAI/Musk case. Anyone with knowledge have an opinion on the veracity of Musk's claims?
1 Robot an hour from Figure now being produced
A really weird part of human history, I think if Robotics come mainstream and say this year is the iPhone 1, I won't consider it until say the iPhone 8.
Assuming a lot will go wrong, the hardware will be an issue interesting to see scaling starting to happen to target factory work, severe bottle necks in production with one an hour for now.
Honestly I don't feel they're going to be capable or reliable enough for a few years, inevitably something will go wrong at some point and it'll be a PR nightmare.
A really weird part of human history, I think if Robotics come mainstream and say this year is the iPhone 1, I won't consider it until say the iPhone 8.
Assuming a lot will go wrong, the hardware will be an issue interesting to see scaling starting to happen to target factory work, severe bottle necks in production with one an hour for now.
Honestly I don't feel they're going to be capable or reliable enough for a few years, inevitably something will go wrong at some point and it'll be a PR nightmare.
Seeing a lot of negative articles about AI creep into the mainstream media the last week or so, thoroughly enjoying it.
WSJ reported yesterday that OpenAI missed on both revenue and user targets for last year. Github Copilot will be going to per-token billing in June. Could be that users will start to see how much AI actually costs to run. If the Copilot sub is any indication then people are going to be pissed.
wsj link - if it works
wsj link - if it works
The amount of people who should know better casually giving unrestricted access to ai agents is fecking terrifying.
Fortunately open source models are getting very good, and very efficient. But yeah I'm already feeling the pinch....
That's cool, but when can I get a C.H.A.P.P.I.E?
Anthropic partners with SpaceX.
How does this make sense considering that SpaceX owns xAI?
Or it is them joining cause both Elon and Dario hate Sam?
Both probably, the new data center xAI will have I'm assuming has so much capacity they're happy to sell compute, the same goes for Google renting out their TPUs to Anthropic
I think Google owns 15% of Anthropic.
xAI indeed has a lot of compute. Maybe having been unable to crack software, Musk will focus on what he knows best, engineering projects and convert xAI division to cloud renting platform.
Mozilla article on Mythos:
https://hacks.mozilla.org/2026/05/behind-the-scenes-hardening-firefox/
https://hacks.mozilla.org/2026/05/behind-the-scenes-hardening-firefox/
US science funding is being cut across the board and funneled through the "Genisis Misson" at DoE. All funding proposals through the NSF are being stopped and you have to write a new proposal through the DoE, with no guidance, and basically make up lies about how you will shove LLMs into your project. Again....with no guidance offered on how to create proposals beyond vague goals written in a report by a LLM.
Anyone here using Openclaw? What do you use it for?
Claude is annoyingly inconsistent. With 4 different accounts it's not responding. When it responds it has great output but the system feels very unstable overall.
Are you using the free or paid version?
Free.
Well, there's your answer. The free tier is running on whatever spare capacity is left over after paying subscribers. I think the instability is a feature not a bug.
Figures package sorting robot doing an 8 hour work shift
Guessing Factories see limited use this year
Guessing Factories see limited use this year
I've been using Hermes agent for over a month..... Skip openclaw imho