TechThoughts

- @BronGondwana fastmail.com/blog/not-writt…

"In a world where there’s enough AI capability to process the entire web and rewrite every page to remove something, the cost of “changing history” is much reduced, so we can expect more of it."

"In the novel 1984, the “Ministry of Truth” has a whole massive department which rewrites history." 1/2

On this day in 1984 this was Apple on “1984”. It was an ad that defined the company and an era. It was anti-big company, anti-big government and pro-individual and pro-privacy. It set the ethos of early computer users. It was another time and place.

We believe our work may be one of the first instances where members of the public have collectively directed the behavior of an AI through an online deliberation process. You can read more about our research in our blog post here: anthropic.com/index/collecti…

To those who say: "but closed source products have billions of investments behind them" I reply: In the mid-90a, Microsoft and Sun Microsystems battled to provide the software infrastructure of the internet, MS with WinNT+IIT+ASP+IE, Sun with Solaris+httpd+Java+Netscape. They both lost out to Linux+Apache+MySQL+PHP+JavaScript+etc. Infrastructure software platforms want to be open source.

OMW to the courthouse this morning for #USvGoogle. Satya Nadella is expected to testify today. I find CEOs are often reluctant to acknowledge the impact of anticompetitive conduct by their competitors because they want to project strength & confidence to their own shareholders...

@OwainEvans_UK But the real kicker from Owain's result is that it implies that hallucinations come from incredibly sharp missing gaps in training data. Implies LLMs are like interactive books They have the info they have... and nothing else Little-to-no deductive logic Good for safety!

1/4: Today in Axios, AIPI released our AI legislative proposal report card. We scored 9 different proposals on whether they achieve a range of priorities for regulating AI that are supported by the public according to the YouGov poll we released Tuesday. We scored policies according to whether they: - Adapt to AI advancements (67% support) - Discourage high-risk AI deployment (66% support) - Reduce proliferation of dangerous AI (86% support) - Affect model training (79% support) - Restrict capability escalation (67% support) Address current vulnerabilities and dangers (82% support) - Prevent or delay superintelligent AI (76% support) See the attached image for how each policy performs according to each criterion

AI systems are fast becoming a basic infrastructure. Historically, basic infrastructure always ends up being open source (think of the software infra of the internet, including Linux, Apache, JavaScript and browser engines, etc) It's the only way to make it reliable, secure, and customizable. Future AI assistants will mediate everyone's interaction with the digital world. This is way too foundational and powerful to be proprietary. In the future, AI assistants will constitute the repository of all human knowledge. If we want it to represent all the world's knowledge and culture, the training and fine-tuning will have to be crowd sourced. Just like Wikipedia.

Remember the signatories of the @FLIxrisk open letter, which called for a pause on advanced AI development? According to a new paper, "Why They're Worried," their motivation to sign had nothing to do with X-risk. Their concerns were NOT centered on "Human Extinction" at all. 1

Interesting poll results by @DanielColson6’s new AI Policy Institute. 82% of US voters don’t trust tech executives to self-regulate on AI, 72% would support slowing down development. Looks like it’s time for governments to step in and ban AGI development in the private sector?

In the last few months, the UK has shown great ambition to lead the world in mitigating risks from AI. But what does this mean in practice? A 🧵. 1/5

IN CASE YOU MISSED IT: All of Apple, WhatsApp & Signal will be pulling their products out of the UK market because of proposed Government surveillance obligations. Yes, really.

One like = one spicy take about AI policy.

@jackclarkSF Import AI 310 twitter.us13.list-manage.com/track/click?u=…

“this paper is a great example of how I expect AI to begin changing the world in a quiet and significant way - all around us, things will become quietly more efficient and imbued with certain sub-sentient agentic intelligences, diligently working away in the service of humanity.”

i was also surprised at how accurately the model predicted my current stated interest (after lots of iterations / trial & error) from decade-old journal entries this made me wonder that maybe this path was actually already seeded long ago in my psyche twitter.com/michellehuang4…

The internet *always* carried "sensitive" information - love letters, private discussions of health issues, political plans - but it wasn't until investors set their sights on *commerce* that the issue of data privacy came to the fore. 18/

Due to the number of people who ticked the box, GameStation claims believes as many as 88 percent of people do not read the terms and conditions of a Web site before they make a purchase."

"A computer game retailer revealed that it legally owns the souls of thousands of online shoppers, thanks to a clause in the terms and conditions agreed to by online shoppers...