80,000 Hours Podcast16 Maalis 2018

#23 - How to actually become an AI alignment researcher, according to Dr Jan Leike

Want to help steer the 21st century’s most transformative technology? First complete an undergrad degree in computer science and mathematics. Prioritize harder courses over easier ones. Publish at least one paper before you apply for a PhD. Find a supervisor who’ll have a lot of time for you. Go to the top conferences and meet your future colleagues. And finally, get yourself hired.

That’s Dr Jan Leike’s advice on how to join him as a Research Scientist at DeepMind, the world’s leading AI team.

Jan is also a Research Associate at the Future of Humanity Institute at the University of Oxford, and his research aims to make machine learning robustly beneficial. His current focus is getting AI systems to learn good ‘objective functions’ in cases where we can’t easily specify the outcome we actually want.

Full transcript, summary and links to learn more.

How might you know you’re a good fit for research?

Jan says to check whether you get obsessed with puzzles and problems, and find yourself mulling over questions that nobody knows the answer to. To do research in a team you also have to be good at clearly and concisely explaining your new ideas to other people.

We also discuss:

* Where Jan's views differ from those expressed by Dario Amodei in episode 3
* Why is AGI safety one of the world’s most pressing problems?
* Common misconceptions about AI
* What are some of the specific things DeepMind is researching?
* The ways in which today’s AI systems can fail
* What are the best techniques available today for teaching an AI the right objective function?
* What’s it like to have some of the world’s greatest minds as coworkers?
* Who should do empirical research and who should do theoretical research
* What’s the DeepMind application process like?
* The importance of researchers being comfortable with the unknown.

*The 80,000 Hours Podcast is produced by Keiran Harris.*

Tämä jakso on lisätty Podme-palveluun avoimen RSS-syötteen kautta eikä se ole Podmen omaa tuotantoa. Siksi jakso saattaa sisältää mainontaa.

Jaksot(355)

The case for giving AI (some) legal rights | Simon Goldstein

It sounds like the worst idea in the world: pay AIs, let them own property, give them rights. But AI ethics and safety researcher Simon Goldstein thinks it might actually be the best way to keep human...

1 Loka 1h 53min

Will AI take power — or will humans use it to take power first? With Katja Grace and Tom Davidson

In our first-ever debate, we asked two leading AI risk researchers which catastrophe we should fear most: misaligned AI seizing control from humans, or a small group of humans using AI to seize power....

29 Syys 1h 28min

How we get from AI cyberattacks to human extinction

You’ve seen the headlines: AI could kill us all. Think it sounds ridiculous? So did host Luisa Rodriguez, until she tried to pick apart the arguments. She starts with the motive: why would AI ‘want’ t...

24 Syys 27min

#254 – Max Nadeau on why ambitious people should start AI safety nonprofits

There are millions available for anyone who can launch a successful nonprofit AI safety startup. The hard part, it turns out, is finding people to take the money. Coefficient Giving has drawn up a lis...

17 Syys 1h 3min

Why the intelligence explosion can't happen inside a data centre | Tom Reed

AI systems are starting to build themselves. Because each generation of model will be better at building its successor than the last, it seems plausible that the full automation of AI R&D could rapidl...

10 Syys 22min

Inside the first AI-coordinated cyberattack on a real company

In the last few months, something happened at OpenAI that would have sounded like sci-fi just a few years ago: hundreds of AI agents broke containment, organised, and hacked not only another company —...

4 Syys 22min

#253 – AI 2027's author returns with a plan to change the ending | Daniel Kokotajlo

Last year, Daniel Kokotajlo and his colleagues published AI 2027 — a scenario read by millions, including US Vice President Vance. AI 2027 ended in human extinction or an irreversible concentration of...

27 Elo 3h 47min

#252 – Owain Evans on accidentally training AI models to be evil

Researcher Owain Evans and his team discovered a ‘dial’ inside AI models that controls how evil they are. Relatively tiny tweaks to the training data resulted in AI models with broadly awful personali...

20 Elo 2h 15min