GOTO 2018 • On the Road to Artificial General Intelligence • Danny Lange

Share it with your friends Like

Thanks! Share it with your friends!


This presentation was recorded at GOTO Copenhagen 2018. #gotocon #gotocph

Danny Lange - VP of AI and ML at Unity Technologies and previously led innovative ML teams at Uber, AWS and Microsoft

Join this session to discuss the role of intelligence in biological evolution and learning. Danny Lange will demonstrate why a game engine is the perfect virtual biodome for AI’s evolution. Attendees will recognize how the scale and speed of simulations is changing the game of AI while learning about new developments in reinforcement [...]

Download slides and read the full abstract here:


Aditya Devgan says:

One of the best talk on AGI

Micheal Bee says:

This is very helpful. I learned a lot. I'm looking forward to analyzing your GitHub code. Cheers! BTW. Is there a talk on connecting functional modules for general AI?

Gogul Ilango says:

This might be the start of real MARVEL's VISION character! Amazing talk.

stivstivsti says:

I don't expect serious AI results soon. All real-word bio organisms have much much more neurons then modern GPU cards. Most of the examples presented were available 10 years ago already:

Pixel Pete says:

so ugh whats this guy do at unity… lol

Thor Mygind says:

The Danish accent was pretty painful, otherwise interesting.

Jozsef Hegedus says:

6.30 … entropy… energy … energy flow … yes, from the Sun … this is basically saying why life exist ? difficult question… entropy is defined only for equilibrium systems, what is entropy ? missing information, what is energy ? size of the information storing capacity of a system – I think I go with Susskind on this

– sorry Danny … I don't buy this, stop watching it at 6.30 and watch Susskind instead … don't wast your time with mambo jambo


take a shot every time he says "yeah"

Sami Alhasan says:

Good Lecture about AGI …. I like it

Nicholas Hauge says:

Simply incredible

Hugues. Freesoul says:

I Loved the intrinsic reward idea, just brillant. The incremental difficulty to achieve best results too. Thanks for this amazing presentation.

Greg Wills says:

Hmm, how do teach object permency in an intelligence that requires repeated simulation in order to learn? Can that be done without making the intelligence aware of the simulation, or even with it? I understand the biological model is a nice road map but the biological model needs a variation of intelligence to survive and often creates intelligences that are unstable, especially in unstable environments.

Ayyy LMAO says:

Just one sentence,

Please make it safe, and decentralized. I'm ready to invest huge money to ensure my kids future

Kittywampus Drums says:

Get involved, SingularityNET. Similar discussion at:

Daniel Tule says:

I am of the exact opinion as this man. i did not understand why nobody else was thinking along these lines. I was looking forward to solving the Ai problem with this "novel" reasoning when I graduate in 2020. Too bad i was beaten to it but great at the same time.

HampTamp says:

#PraiseTheCameraman lol

Joke aside, this is a really interesting talk. I don't want it to stop. Should've been 5 hours long at least. Now I'm really curious about AI

C.C. ORR says:

only 2,743 views. The whole of youtube for the past month or so. the most amazing anything ive ever seen pondering on the implications for humanity and 2,743 views? eh?

Kevin Wildberger says:

These guys should study how slime molds make decisions. Are these guys mimicing human intelligence? Or something else?

Anders flortjärn says:

Best AGI talks on YouTube at the moment

zapupyrla says:

The last part is very thought provoking. So how we as humans made this leap without increasing the power of our brain? You said it is because of communiction and knowledge transfer. So if AI learn how to transfer knowledge, then it can evolve faster. But when AI learn to transfer knowledge, immediately arises the problem of whom to to trust in this transfer. Then some learning about the trust should emerge some tribes or other "cultural" phenomena among AI! Another thing is that when you transfer knowledge between trusted units, you do not challenge this knowledge, so this is how Traditions arise. Then, as in humankind, you need intelligent members, "Artists" and "Creatives" who challenge the traditions and culture to find possible jumps to a better local point on the optimization landscape… Wow!

Frank Staheli says:

Fascinating how a video game can teach itself how to become more successful at achieving an assigned goal!

Write a comment