While @BlueDotImpact HQ is not moving to SF until next year, a bit of the BlueDot spirit arrives @vivariumsf this Saturday. Join us and accelerate the right kinds of things.

the future is extremely promising, but it has some advanced dangers. ​so let's def/acc -- accelerate the defenses to big risks, so we get the good future. hackathon nov 22 at @vivariumsf. sponsored by @BlueDotImpact (others to be announced soon). https://t.co/VPd0vyylhl

running what will be the world's largest ever def/acc hackathon with @joshua_muthu, @AniketChak19034 and prince on non 21-23. online and in-person in london (with more cities coming soon). you should apply and help us flip the offense-defense balance on cyber, bio and control.

@certainforest @OpenAI Check out their work:

Congrats to our alumni @certainforest and Charles Ye for winning @OpenAI's Red-Teaming Challenge (out of 600 entries!) 🎉 They exposed how "safe" AI can be tricked into breaking its own rules (89-95% success rate). Kudos to their work on making AI safer!

£10k if you find me our next tech lead - I think this is a very cool role (I'm covering for parts of it atm) at a very cool org (we just raised 25M). DMs open :)

Last week we co-hosted our first hackathon with @join_ef This week we're announcing a $1M fund for our new AGI Strategy Course Next week the first funded project enters the world We move fast at @BlueDotImpact You want in? Well, are you an LBE? 🧵

We’re putting $1M behind people who want to ship solutions to AI risks. You’ve seen the rate of AI progress. You know you need to do something. But you’re uncertain. What should you build? What’s the most impactful thing you can do?

AI models are becoming ever more capable, enabling risks at unprecedented scale and complexity. That's why we're teaming up with @EntrepreneursFirst and @WorkshopLabs to host our first AI Security hackathon! 2k GBP in prizes across winning teams.

I have it on good authority that if you register interest in our new AGI Strategy Course today, we'll invite you to apply to our first test run starting Aug 25th.

Big week for more people paying attention to what’s happening with AI: a cover story + 2 briefings in The Economist! Didn’t feel this traction earlier in the year when we were building @BlueDotImpact’s Future of AI course. Lots of work ahead - but feels good seeing the momentum.

🚨 We're hiring at @BlueDotImpact to build the AI governance pipeline. Your mission: Figure out what governance is needed to make AI go well, then build the workforce to make it happen. Imo, this is a top 1% role if you want influence on AI's trajectory 🧵

Want to know more? Read our full blog post:

Ultimately, we don’t have enough evidence to be sure whether AGI is decades or just a couple of years away. We may have very little time to mitigate all the risks that very powerful AI could pose.

It’s also not clear whether an intelligence explosion could happen in reality. Even very smart automated AI researchers could run into shortages of data and compute.

Even though AIs are closing in on expert level at abstract reasoning tasks, robotics is lagging. AGI-skeptics point to Moravec’s paradox – the fact that tasks with a long evolutionary history are harder for AIs.

However, not everyone is convinced. Some people think we’ve got a lot further to go before we reach AGI than extrapolating benchmarks would suggest. They point out that benchmarks tell us a lot about easily-measurable tasks – but the real world is much messier.

Even though there’s still a wide spectrum of opinion, it’s far easier to find examples of people shortening their timelines than the other way around. Take forecasting platform Metaculus, where the predicted timeline to AGI has dropped by over two decades since 2022.

AI companies are piling massive resources into training more powerful models, and research suggests that this won’t slow down before 2030.

If we can get AIs to help us with AI research, things could go even faster. We might even kick off an “intelligence explosion”, where each generation of AI builds its own successor.