Ramez Naam

Climate and clean energy investor. Author of 5 books. Energy & Environment co-chair @SingularityU. Trying to build a better world.

[Ramez Naam @ramez]

Despite this election, I remain an optimist about America and the world. Humanity will continue to produce new ideas and new innovations to improve our lives. Good people will continue to come together to improve the world. And the political tide will turn. We'll make it so.

[Ramez Naam @ramez]

@Higgs_David_J That's good to hear. A lot is lost in translation. A lot of very informed AI people are surprised when I show them these numbers.

[Ramez Naam @ramez]

Good find by Toby. AIs are far less capable on real world AI research tasks than ECI x METR or AI 2027 predict. Data from Google Argon (adding on to OpenAI):

Gemini 4 actual: ~15 minutes
ECI --> METR Prediction: ~360 minutes (6 hrs)
AI 2027 prediction: ~1,000 minutes (18 hrs)

[Toby Ord @tobyordoxford]

Fortunately, that has arrived. GDM ran a very similar analysis on their internal use of AI for AI R&D and published their graph on page 78 of their Gemini 4 Safety Report:

[roon @tszzl]

this is just wrong: these vulns were found by defenders using Daybreak rather than attackers. It's just evidence Glasswing didn't go far enough, possibly due to capacity constraints. It's also very likely that March Mythos was not as good as Sol/Astra of September

[Ramez Naam @ramez]

@KelseyTuoc ❤️

[Ramez Naam @ramez]

That is certainly not how it's communicated. METR purports to be estimating the length of real engineering tasks. AI2027 uses length of coding task to estimate when a real world event - a runaway singularity - could occur. It's very relevant to point out that their numbers don't reflect real world AI research.

[Ramez Naam @ramez]

@Birdyword @bswud @s8mb Congrats!

[Ramez Naam @ramez]

@Higgs_David_J Yeah, that's a hard one and somewhat subjective. The best intel I could get suggested that 6 - 8 weeks before public release was roughly when it was broadly in use internally. So the current model has ECI shifted 60 days in advance. I'm open to better ideas if you have them.

[Ramez Naam @ramez]

@alethiawiz It's not clear it's doubling. That would be the straight orange dashed line. But linear and sub-linear trends of various sorts fit better.

[Jane Flegal @JaneAFlegal]

Making it so that data centers can get powered by the grid relative to a counterfactual where they get powered by off grid gas (should be the #1
goal for climate advocates btw) does not make this a “pro data center bill” at all

[Ramez Naam @ramez]

@StatisticUrban The Peripheral
Westworld
The Lazarus Project
The OA

[JD Ross @justindross]

There were 4 CVSS-10 security incidents over the past 24 hours. One of them is the worst incident I’ve seen in my life.

So we created an auto-SRE tool that polls for security incidents, identifies which ones we're in trouble for, and automatically fixes them.

[Ramez Naam @ramez]

@dok2001 @ThePHPF Thank you for your service!

[Ramez Naam @ramez]

@TheDavidSJ @BogdanIonutCir2 Yep. That's a real issue. Nevertheless, real world data has a different level of authority than synthetic benchmarks.

[Jane Flegal @JaneAFlegal]

GET THEM ON THE GRID

[Joel Grus 🤠 @joelgrus]

she's a 10 but that was supposed to be a random 384-bit number

[Matthew Prince 🌥 @eastdakota]

Some stats on #php2shell. In the 10 hours before it was publicly announced, across the @Cloudflare network we saw thousands of attempts to exploit it. That jumped to tens of thousands in the 45 minutes post announcement, hundreds of thousands within 2 hours, and millions now.

[rohit @krishnanrohit]

Some context on the trend of cyber incidents against banks

[Shakeel @ShakeelHashim]

News! OpenAI's written to Gov. Gavin Newsom recommending a bunch of amendments to strengthen SB 53. Includes a bunch of stuff on AI R&D and shutdown controls.

[Alex Imas @alexolegimas]

It me

[Ramez Naam @ramez]

@dens You were a pioneer. You set a lot in motion. Thank you.

[Anthony Ronning @anthonyronning]

If you need defenders access right now to open source cyber models, hosted securely and privately, and you can’t or won’t go to the major labs for it, reach out.

We’re evaluating something privately for people we know at @TryMapleAI.

Already deployed but not publicly launched.