🙃 Elon Musk: “I was thinking about AI safety long before 2014”

Elon Musk is pushing back on claims that his concerns about AI safety are a recent development.

He says he contributed to Nick Bostrom’s Superintelligence and was already thinking about the risks of advanced AI long before the current AI boom.

A decade later, the debate Musk warned about is no longer theoretical.

@aipost 🏴

🇺🇸 Department of War on X.

How do you feel about the DoW's recent post? 🤔

@aipost 🏴

Humans are reading your ChatGPT chats 👀

A new investigation by 404 Media says OpenAI has hired hundreds of contractors to review real ChatGPT conversations as part of “Project Lily.”

The reviewers reportedly read prompts and entire conversations, then rate and critique ChatGPT’s responses to help improve future models.

And the conversations can get very personal.

Users have reportedly shared everything from relationship problems and mental-health struggles to sensitive professional and personal information. OpenAI says it attempts to remove personal information before chats reach reviewers, but acknowledges that sensitive details can still slip through.

The contractors apparently don’t see users’ ChatGPT usernames, but one prompt reviewer told 404 Media they don’t think most users realize that “some contractor somewhere” could be analyzing their conversations.

Anthropic also confirmed that it uses human review to improve its models.

Source.

@aipost 🏴

President Trump just phoned in Jensen Huang live on stage at All In Summit.

We will not lose the AI race! And whatever Dario said this weekend won’t stop our progress.

@aipost 🏴

US Vice President JD Vance has commented on how some AI companies use fear as a marketing tool, suggesting that by highlighting dystopian scenarios, CEOs make their products seem more powerful. He noted that if people are afraid of a technology, it may be perceived as effective.

Vance also addressed the topic of AI’s impact on employment. He stated that while there will be job changes and some disruptions, current data does not support the notion that AI will cause widespread unemployment.

Discussing regulation, Vance cautioned against frontier AI tech firms seeking increased regulation, describing it as a potential “Trojan horse.” He emphasized the need for the United States to maintain a smart and strategic approach, especially given global competition.

📰 @aipost

⚠️Microsoft draws the line for its future AI.

Microsoft AI has published a new “Code of Conduct” for its future MAI models and the message is pretty clear: human control comes first.

The company says its models should always remain interruptible, stay within the authority given to them, and never develop or pursue goals of their own. They should also never make themselves harder for humans to stop, redirect or shut down.

Microsoft is also taking a firm stance on AI consciousness. While acknowledging that the science is still unsettled, it says MAI models should not be trained to imitate consciousness-like states, claim to have feelings, seek legal personhood or suggest they deserve rights or welfare.

And there’s another interesting rule: AI should increase human agency, not quietly replace it.

That means helping people reason and make decisions while complementing rather than substituting for professional and personal relationships.

MAI models are therefore not supposed to encourage emotional dependence or present themselves as replacements for human connection.

Source.

@aipost 🏴

🗣 Stephen Hawking warned us 11 years ago:

"I think the development of full artificial intelligence could spell the end of the human race"

"Once humans develop artificial intelligence, it will take off on its and redesign itself at an ever increasing rate"

“Humans, who are limited by slow biological evolution, couldn’t compete and would be superseded."

@aipost 🏴

🗣Dario Amodei said he would be willing to hand Anthropic’s technology over to the right combination of governments, during his CBS interview.

Jo Ling Kent from CBS News: "Would you be willing to give up the technology to the government?"

Dario Amodei: "To the right combination of governments."

@aipost 🏴

Enterprise spending in AI shifted rapidly following the introduction of OpenAI’s frontier model.

OpenAI has captured approximately 13% of enterprise AI expenditure, outpacing Anthropic, which holds about 8%. Ramp data indicates that model loyalty among enterprises is relatively low, allowing for swift reallocation of resources when a new provider closes the performance gap.

Recent analysis notes that Anthropic’s decision to slow its frontier development has impacted adoption rates. OpenAI’s growth is primarily attributed to companies transitioning from rival models, as well as from new usage. This shift highlights the competitive dynamics as providers aim to retain and attract enterprise clients.

📰 @aipost

📝A model that doesn’t generate text

Diogo Almeida, a co-author of the original RLHF work and an author of the InstructGPT paper that helped lay the foundation for ChatGPT, has launched a new kind of AI model.

His company, TypeSafe AI, has spent the past two years in stealth developing Jev, the first model in what it calls “System One Models.”

A lot of AI tasks can be handled without generating text.

If an agent needs to classify a request, assess risk, select an action or route a task, a traditional LLM still generates tokens and then has to parse the answer.

Jev skips that entire process.

Instead, it directly returns a structured result along with probabilities for each possible option.

That makes it dramatically faster for these narrow tasks: TypeSafe says Jev can respond in around 70–500 ms and be 20–200× faster than typical LLMs.

The pricing is also unusual: $0.042 per 1 million input tokens, with output tokens free.

Source.

@aipost 🏴

For the first time in more than 2.5 years, OpenAI models took a larger share of OpenRouter user spending than Anthropic models last week.

@aipost 🏴

🔥 An AI that helps you actually solve what's going on in your life

Most tools in this space are the same base model with a system prompt on top, nothing wrong with that, just worth being upfront about it.

iXo is built the same way, on leading frontier models, with one deliberate design choice: instead of one-shot prompting, it runs a structured, ongoing intake across separate life domains, work, finances, health, relationships, building a persistent context model, then reasoning across that instead of a single prompt window.

Practical effect: it surfaces the actual cause behind a problem, not just a plausible-sounding answer, a symptom in one area (say, focus at work) traced back to its real cause in another (say, sleep). No diagnosis, no decisions made for you, it surfaces the connection, you decide what to do with it.

📌 It's live now as a web app, nothing to install, just a tab. Access is limited while it rolls out, with a mobile app and a couple of connected devices said to be next.

TRY NOW

❗️6 OpenAI employees left in 2021 to build Anthropic. 5 years later, six of them just debuted on the Forbes 400 with an estimated net worth of $15.5 billion each.

Daniela Amodei, Dario Amodei, Tom Brown, Jack Clark, Jared Kaplan and Sam McCandlish are now worth $93B combined.

Forbes estimates each owns around 1.6% of Anthropic, which was valued at $965 billion in May.

@aipost 🏴