Superintelligence Could Arrive in 1 to 2 Years | Dewi Erwan artwork

Superintelligence Could Arrive in 1 to 2 Years | Dewi Erwan

MTS

November 10, 2026

BlueDot Impact co-founder and CEO Dewi Erwan breaks down the bottleneck in AI safety talent, the prospect of superintelligence landing within two years, and why the field desperately needs founders and government advocates rather than more machine learning researchers.

0:00

Dewi Erwan

If we are on track to have superintelligence in a year or two, this is a terrifying situation.

And I would prefer for us to have more time to be able to figure out what do we actually want to happen with this technology before the most transformative thing that's ever happened to our species just lands on our doorstep with very little notice. And also trying to get the AI companies, who for many, many years have been telling us that the technologies they're building could kill everyone, getting them to be like to reflect on well, why are you still racing to build the technology, which you have been claiming since you have been founded could kill everyone.

0:38

SPEAKER_2

And we are back. We're live with Dewi Erwan, who's co-founder and CEO of BlueDot Impact, which helps people get into careers in AI safety. So Dewi, welcome to MTS.

0:50

Dewi Erwan

Thanks so much for having me.

0:51

SPEAKER_2

So how did BlueDot come to be in the first place?

0:54

Dewi Erwan

Yeah, so back when I was at university, I got very worried about a bunch of different problems. So growing up initially, I was very passionate about climate change, and I wanted to work on nuclear power and a bunch of renewables. So I went to uni to do engineering. At uni, I got quite worried about factory farming. I thought, okay, I want to build like cultivated meat or alternative proteins to try and change the way that we produce food.

And then around 2018, 2019, I learnt more about AI and bio, and learnt about the various ways that biology could be used to cause harm. And every single time I learnt about these different problems, I found myself being very, very blocked on what do I actually do? How do I help as an individual to solve these huge global problems?

And it was infuriating to me that it was so difficult to actually do anything useful. So we then decided after graduating, I moved to Cambridge and started trying to help solve this problem for other people. It's like building talent infrastructure for enabling people to learn more about these problems, figure out how they can contribute. And then, since then, we've just been scaling up since.

2:08

SPEAKER_3

I'm really curious about how long does it take to move somebody, give them the necessary context and move them through the sort of pipeline, like from when they show initial interest, maybe as like young university grads or what have you.

2:20

SPEAKER_2

Well, it's got to depend on the person.

2:21

SPEAKER_3

Yeah, it probably depends on the person.

2:22

SPEAKER_2

I'm also curious about the difference in how long it takes for different kinds of people.

2:26

SPEAKER_3

Just really curious about the candidate space.

2:28

SPEAKER_2

Smart college student versus mid-career software engineer or something.

2:32

Dewi Erwan

I think the, yeah, there's a few inputs here. One is how much infrastructure actually exists to accelerate people, and then the second is how much can people do autonomously? So you can now already get the first few tens of hours of learning through, for example, the course or starting to do a bunch of other reading books.

But if you actually want to get much more context on the field, it is still very, very difficult to do this. If you want to get into one of these fellowships, they run twice, three times a year, so you have to wait five or six months until you can actually start MTS or start GovAI or something like this. So it often takes people one to two years to actually get into the field. So one of the things that we're now doing is trying to scale up our grant making so that we can actually give people money to quit their jobs or give people money to quit university and actually start working full time on learning way more about the problems and how they could contribute.

And we're also building a campus near here as a place where people can go. Like, you know, we have a conversation with them, and they're like, yeah, I learned about AI last week. I got really, really worried. I sent them a few resources, have another chat with them two days later. They're like, you know, seem really, really smart. I'm like, cool. We can give you $50,000 grants, enable you to leave your current role and start learning way more about the field, build your own understanding, build your own takes about what is the problem, how could they contribute, and then we fly them out and help them pivot as quickly as possible.

Read the other 16 minutes of this episode here.

$15 gets you this episode and 99 more. Credits never expire. After checkout, one link brings you back here to read it.

Not what you expected? Full refund within 14 days.

Thousands of transcripts fetched by people building searchable podcast archives

Fetch transcripts with code

The same credits work from a script or an AI agent. The demo key returns a sample episode in full, no card needed:

request
curl -H "x-api-key: pt_demo" \
  https://spoken.md/transcripts/1000651996090

Markdown with the speakers named, for your notes, your knowledge base, or anything that makes HTTP calls.

From $0.10 per transcript. No subscription. Credits never expire. Prices exclude VAT, added at checkout for EU customers. Not what you expected? Email us within 14 days with 20 or fewer credits used and we refund the pack in full.

Using your own key:

request
curl -H "x-api-key: YOUR_KEY" \
  https://spoken.md/transcripts/1000800989367