Will AI Gatekeepers Tell You What You Can Build? - Mark Zuckerberg
Ғылым және технология
Full Episode: • Mark Zuckerberg - Llam...
Apple Podcasts: podcasts.apple.com/us/podcast...
Spotify: open.spotify.com/episode/6Lbs...
Transcript: www.dwarkeshpatel.com/p/mark-...
Me on Twitter: / dwarkesh_sp
Пікірлер: 45
Notably for cybersecurity, since you can't Fine-tune a GPT-4 for cybersecurity because it can be used with harmful intentions, it's even worse for Google Gemini API. LLaMA 3 really is a big deal for the open-source community.
Gatekeeper of social media worried about gatekeepers of AI
@Nasser-bp6qf
16 күн бұрын
what are they gatekeeping?
Is the Meta VR store not a gatekeeper? There's not even a formal application process to get your app onto the store. It's by invite only. I've seen numerous VR developers move to other development fields, because there isn't access to the market. Meta is gatekeeping it and basically defining what you can and can't build.
@thems_the_brakes
24 күн бұрын
Of course he’s talking out of both sides of his mouth
@RR-et6zp
22 күн бұрын
rules for thee not for me
When lack of oversight meets profitability
@michahcc
17 күн бұрын
Resonance of Josh hawley
AI is being gatekeepers kept like crazy! You can’t even use it…you can only use stupid app toy versions.lol
How about a open source social networking platform that replaces all the SN giants and forces a lot less ads (just enough for running cost)?
@kushagra1064
21 күн бұрын
good luck with that
@prakash27502
21 күн бұрын
Mark Zuck is not talking his bread & butter revenue source Social networking to take to open source. he is also not talking about sharing/open sourcing the people's data that FB collected for free. I selective altruism. :-)
@GeorgeMaci
21 күн бұрын
See mastodon that is what’s already happening
@johnswanson217
21 күн бұрын
SN is not just a code. Code is about 10% of their cost. The cost is all about compute and networking resources.
@prakash27502
21 күн бұрын
@@johnswanson217 AI too has high cost of compute and inference
Meta killed many companies by closing the Events and Groups APIs and has been one of the worst gatekeepers in the big tech ecosystem historically. Please interview Jay Graber from Bluesky so more people can start to understand what the future can look like.
@Macatho
18 күн бұрын
Facebook was at least open in the beginning but when every other company started closing their doors for free access to all the data on the platform... Facebook had to do it as well. It's crazzy that Facebook is magnitudes more open source than OPEN-AI...
@mikestaub
18 күн бұрын
@@Macatho Facebook only seems to open source when it's in their corporate interest. There are ways of commodifying competition similar to why Google open-sourced Kubernetes to hurt AWS's lead.
@Macatho
17 күн бұрын
@@mikestaub Perhaps, does not change anything though.
@mikestaub
16 күн бұрын
@@Macatho We don't know that as we don't have the counterfactual. If Facebook was still an open platform it may have had profound implications on society for all we know.
@Macatho
14 күн бұрын
@@mikestaub perhaps, and perhaps not.
No nie 6 dni temu.
Wow, I never thought I'd join Z's side on anything haha Now, go make Facebook into a positive force for the West's unity.
Never thought I'd say it, but long live Zuck! 😅
@GM-xk1nw
23 күн бұрын
Can't say the same for all the people who died because of him
@the_original_dude
22 күн бұрын
@@GM-xk1nw nobody died because of him
@M-dv1yj
20 күн бұрын
U ever considers this is his pr attempt to stay his execution when things all fall down??
One thing I found interesting and slightly concerning in this conversation is that Mark really didn't have an aswer to some of the very important questions and several times it seemed as if you were asking questions he hadn't even considered and was just winging the answer. Like the question of "what if not all issues are like cyber security where an advanced defender wins against a comparatively primitive malicious actor", which to me is a super important one. By the second law of thermodynamics it's so much easier to destroy something than it is to build it, and to me cyber security seems like an exception to the rule not the norm. Give a bad actor nuclear weapons, a bio-engieneered virus, a easily mass producible deadly chemical and no amount of sophisticated AI defence is stopping them before a LOT of damage has been caused. Focusing on counteracting mis-/disinformation is all well and good, but the fact that he hadn't seemed to reflect over such a fundamental issue as one of the most important people in the industry is very concerning to me. Good on you for asking the hard questions, cause these guys really need to hear them.
@thems_the_brakes
24 күн бұрын
Very true
@marshallmcluhan33
23 күн бұрын
It's all about perception, America killed hundreds of thousands with nuclear weapons but people generally don't refer to them as "bad actors". Offense and defense are kinda the same just like good and bad actors are the same. I'd rather many people know about this tech before someone sets it off even if it's determined as a just use case by the status quo. Education can go a long way in preventing even unintentional misuse.
@noone-ld7pt
22 күн бұрын
@@marshallmcluhan33 I mean the discussion on moral relativity is definitely one worth having but that's not really what I'm getting at here. I just fundamentally disagree with the statement that offense and defense are the same or even similar in very many cases. Let's take a bio-weapon as an example. If someone develops a virus with the R0 of covid and mortality rate of Ebola and by the time even the best labs in the world has a vaccine the global population could be halved or worse. No amount of bio-offence is changing that. Or better yet, mass release a chemical that depletes the ozone layer at unrecoverable rates. We know that is perfectly doable since we already did it with CFCs, and presumably there are more potent chemical compounds that we have yet to discover. You can have the best CBRN defense in the world, but without an ozone layer we're all toast. Literally. My point is that regardless of who you define as bad actors, on a very fundamental level the more people have access to dangerous tools, the more people get hurt. Both intentionally and unintentionally. And once these models become capable of advanced R&D so many fields suddenly become a lot more dangerous, both directly and indirectly. And I personally think it's terrifying that the people, like Mark, who are pushing the limit of this tech seem so woefully ignorant on questions like this.
@marshallmcluhan33
22 күн бұрын
@@noone-ld7pt I suppose they'll always be tradeoffs and risks. Electricity and combustion engines are very dangerous and hurt the environment on a massive scale but we enjoy their benefits. What can we do to prevent the issues of technical advances?
@nuance9000
17 күн бұрын
What!? That's not the second law of thermodynamics... If anything, the second law of thermodynamics hints at the fickle nature of current AI models. They're both unpredictable (high entropy) and unoriginal (distribution). So it's hard to see the threat of current AI with geopolitics. Heck, our culture of bad passwords is more of a threat than any AI system
for the algo
Hello
He hasn’t answered the question
Mark actually seems based sometimes!
It is a false dilemma to choose between freely distributed AI and a handful of companies having it. Both are awful options. Global agreements and democratic oversight can also be put in place; this must be inserted into the discussion more often. Lets consider the multi-billion-dollar venture of CERN, and that AI is currently of much greater importance.
@LotsOfBologna2
24 күн бұрын
Open source is the best assurance we can possibly have of privacy and anonymity in our use of tech in this time when both are being terribly invaded. There is no justifiable reason to oppose open source.
@ikotsus2448
23 күн бұрын
@@LotsOfBologna2 privacy and anonymity are valid, but also the least of my worries.
Are you people insane? Do you realize there is no such thing as open source to the general public