Derek Nassler @dnassler

Software Developer and Artist Toronto Joined February 2009

Tweets

554
Followers

176
Following

1K
Likes

2K

Derek Nassler @dnassler

3 weeks ago

@beffjezos So deliberately training into the model to disobey its user in certain circumstances when sleeper mode activates? What could go wrong?

0 0 2 66 0

View Details

Derek Nassler @dnassler

2 years ago

@ryangerritsen Shut up! You are just trolling and trying to spread right wing propaganda!

0 0 0 5 0

View Details

@TeamYouTube @damiengwalter @YouTubeCreators @YouTube Seems the YouTube algorithm is not functioning properly and its decisions are not in fact made carefully at all. Please do a human review and undo your mistake.

0 0 3 59 0

View Details

Derek Nassler @dnassler

3 years ago

@post_relativity @Plinz Consider the new book by the physicist @carlorovelli about White Holes where he explains his ideas about Black Hole singularities

0 0 2 41 0

View Details

Derek Nassler @dnassler

3 years ago

@RandolphCarterZ @GaryMarcus Bigger AI models are not a risk to society. Bad humans using smart tools are. Bigger models could be smarter.

0 0 0 17 0

View Details

Derek Nassler @dnassler

3 years ago

@Aceofspades256 @keithfrankish @cathoderayzone @Philip_Goff Why would they not want the discussion to go there necessarily? That could be interesting.

1 0 3 38 0

View Details

Andrew Ng @AndrewYNg

3 years ago

I'd like to respectfully point out that the logic in this argument is based on a flawed model for how scientists think. Scientists don't just take a weighted average of others' opinions to form their own. A good scientist takes as input lots of data, including others' opinions, and then ultimately has to reason, build their own internal model of the world, and draw their conclusions from that model. I give your opinion a lot of weight. And, after having heard many opinions including yours, my internal model tells me that there is essentially no risk of AI human extinction. So I don't follow the logic that because @ylecun or anyone else disagrees with you (and other AI extinctionists, which IMO are in the minority) that they gave your opinion very little weight, unless we think scientists arrive at conclusions by taking a weighted average of what everyone else thinks. Since you've gone against the grain many times in your own career -- often brilliantly so -- I assume you're also familiar with what it feels like to give someone's opinion weight, but then ultimately to draw a different conclusion!

217 579 7K 1.1M 753

View Details

Derek Nassler @dnassler

3 years ago

Some Candid Career Advice newyorker.com/humor/daily-sh…

0 0 0 77 0

View Details

Derek Nassler @dnassler

3 years ago

@alistair___s @erikphoel Thumbs DOWN! Thumbs DOWN! Thumbs DOWN!!!

0 0 1 61 0

View Details

Derek Nassler @dnassler

3 years ago

@s_batzoglou @pmddomingos Imagine if there are different flavours of AI disinformation detection apps with different ideological worldviews…

0 0 1 47 0

View Details

Derek Nassler @dnassler

4 years ago

@firepile I agree that this is absurd and annoying and problematic. If this is an example of what AI Safety people think is a good idea, then imagine what bad AI regulation could look like.

0 0 0 16 0

View Details

Derek Nassler @dnassler

4 years ago

@Grunthosi @Carnage4Life I don’t think that people should believe nothing. People need to know and have some trust in the source and then verify from another independent and trustworthy source. Maybe there we can be tools that help

0 0 0 11 0

View Details

Derek Nassler @dnassler

4 years ago

@SurviveThrive2 @keithfrankish Except that they don’t vet who you are anymore. They only have to verify that you’re payment processor transfers $ to them.

0 0 0 21 0

View Details

Derek Nassler @dnassler

4 years ago

@PerryDahlia_ @dh7net @EMostaque But OpenAI has said that they curated the training data and have said how important doing that is for the end result in terms of quality (and (probably) alignment to some extent as well).

0 0 0 581 0

View Details

Derek Nassler @dnassler

4 years ago

@Kenso_Kay @dh7net @EMostaque Also alignment-wise

0 0 0 30 0

View Details

Pedro Domingos @pmddomingos

4 years ago

In which I rip the AI moratorium letter to shreds: link.medium.com/n2DFddmE8yb

8 16 65 7K 5

View Details

Derek Nassler @dnassler

4 years ago

@Guuber42 @robkhenderson This would make me think my roommate thinks that I might have a poisonous spider on my back except for the crying part. The crying part makes me think the roommate has issues.

0 0 0 39 0

View Details

Derek Nassler @dnassler

4 years ago

@MCorbettHawkins @tegmark No it will not necessarily “provoke” the use of nuclear weapons. Only humans will. This is hyper pessimism. This type of attitude truly makes me sad. The stakes may be high but to give up so early seems very wrong.

1 0 0 10 0

View Details

Derek Nassler @dnassler

4 years ago

@realtimeai But that’s assuming a PEFECT AGI is possible. What if only an approximation of AGI is possible? What if such a thing is moulded by its training data and that training data is biased overwhelmingly towards religious worldview X?

1 0 1 33 0

View Details

Derek Nassler @dnassler

4 years ago

@tegmark The article is just the standard doomsday scenario but human beings NEED to be part of the loop . Everything an AI sights NEEDS to be verified. Suggestions by AI involving changes to the environment shouldn’t be followed blindly!

0 0 1 11 0

View Details

Derek Nassler @dnassler

4 years ago

@tegmark I read the link and it is not convincing me that AGI is necessarily bad. It is very negative about the potential and the examples are poor and have assumptions about human beings allowing the poor decisions of an AI to affect the world negatively such as the ocean acidity example