Will AI Replace Us? With Neil deGrasse Tyson & Matt Ginsberg

Рет қаралды 279,916

Күн бұрын

Is artificial intelligence taking over? Neil deGrasse Tyson and co hosts Chuck Nice and Gary O’Reilly discuss deepfakes, AI hallucinations, and whether AI really is intelligent with software engineer at X, the moonshot factory, Matt Ginsberg.
Have we been using AI in the right way? We explore deepfakes and the technology to verify an image’s legitimacy. Is artificial intelligence really intelligent? We break down what AI can and cannot understand. Does AI understand causality? Or even reality?
What is AI bad at? Could it help us discover objects and phenomena we never knew to look for? We discuss how it could help us explore space or call plays in a football game. Could we one day see AI calling plays instead of coaches? Do two machine coaches cancel eachother out?
Can AI predict the weather? We discuss the concept of chaos and how close we are to artificial general intelligence. Find out about AI’s relationship to the truth and how we can combat people using it to spread misinformation. Finally, could deepfakes become so good that it's the end of the internet?
Thanks to our Patrons Kathleen Kussman, Craig Hamilton, Denis de Oliveira, Jim, Ryan, and Krishna for supporting us this week.
NOTE: StarTalk+ Patrons can watch or listen to this entire episode commercial-free.
Get the NEW Cosmic Queries book (5/5 ⭐s on Amazon!): amzn.to/3dYIEQF
Support us on Patreon: / startalkradio
FOLLOW or SUBSCRIBE to StarTalk:
Twitter: / startalkradio
Facebook: / startalk
Instagram: / startalk
About StarTalk:
Science meets pop culture on StarTalk! Astrophysicist & Hayden Planetarium director Neil deGrasse Tyson, his comic co-hosts, guest celebrities & scientists discuss astronomy, physics, and everything else about life in the universe. Keep Looking Up!
#StarTalk #neildegrassetyson
00:00 - Introduction: Is AI good or Bad?
03:00 - X, Alphabet’s Moonshot Factory
09:16 - Detecting Deepfakes
13:03 - Are we using AI correctly?
16:32 - Will AI make the next deep space discovery?
22:30 - AI vs. Humans
23:40 - AI Calling Plays
34:12 - Can We Predict Natural Disasters?
39:30 - Artificial General Intelligence & Hallucinations
44:17 - Combating Misinformation
48:45 - Is it the End of the Internet?

Пікірлер: 917

@StarTalk 10 ай бұрын

What do you think we'll use AI for as it continues to develop?

@chineseelonmusk9712 10 ай бұрын

Do physical labor most efficiently

@FacesintheStone 10 ай бұрын

We will use it to design manufacturing processes that will create raw materials or final products autonomously. Natural intelligence is something that has gotten us very far, we are able to fly, create vaccines, I think we should learn from natural intelligence before we start creating our own.

@wildgoose419 10 ай бұрын

We should use AI to call out people's BS, unless that AI is designed to do BS, in which case we'll let the non-BS AI duke it out with BS AI, and watch the confrontation with beer, wings, fries, dips, and some occasional celery sticks.

@RoboKubik8 10 ай бұрын

Given enough training data it can do anything. Human brain evolved into what it is now and AI evolves quite similarly. Different training models and approaches will teach them different things.

@KentonJoseph 10 ай бұрын

Map asteroids with threat determination would be good.

@BenD_Bass 10 ай бұрын

At first, I came here to learn and listen to science space stuff. But after like 100 episodes, I'm excited for Chuck. I know Neil and the comment section always pick on him, but that dude is so smart. I don't base smartness on degrees and how much you know. I think not being afraid to ask questions and asking the right questions. He's like good at being curious

@drewparcel1727 10 ай бұрын

He is good at asking the question that we are all thinking and wanting to ask ourselves.

@xtins 10 ай бұрын

Wtf are you talking about? He’s pretty smart, is his ROLE to make it comical … 😅

@loccc88 10 ай бұрын

I agree. I also feel that stand up comedians in general are naturally smart people. It takes a very unique kind of brain to come up with a funny line on the spot.

@MrTekniqs 10 ай бұрын

You understand that education does not equal intelligence. I applaud you.

@JesseJames83 10 ай бұрын

Also here for Chuck

@eric-. 10 ай бұрын

Chuck came PREPARED for this conversation! contagious enthusiasm, as usual. 👍

@Random_user_8472 10 ай бұрын

The honesty, humor and great energy between these guys makes the whole video, everytime again!

@KentonJoseph 10 ай бұрын

Topics and conversation is great and moves topics forward. Also Chuck is great at keeping it lively.

@brian9731 10 ай бұрын

I've rarely if ever seen Chuck so serious on this show as he is in parts of this show.

@dustylong 7 ай бұрын

I liked it a lot 😁

@daniellecolon2281 4 ай бұрын

So serious for real

@B3R34L 10 ай бұрын

Sometimes i wish Startalk would hire a animator to animate in a simple way what some explanations looks like. I've seen it a few times but if all explainers contained animations then i think starttalk could gain so much more followers and views. Just a thought

@frogz 10 ай бұрын

honestly this could be good, i dont want it to be done infographics style though.... if you make something cheap, it will feel cheap

@Bull1the1Great 10 ай бұрын

I always said this . At least now with AI lol, animators have an easier time ...

@jhsounds 10 ай бұрын

Regarding the John Lennon AI story: McCartney has clarified that AI was merely used to isolate the vocal from an existing Lennon demo recording, in order to clean it up for a new song. Lennon's voice was not "synthesized".

@robertadams6606 10 ай бұрын

That's correct his voice was already on the Demo. They just isolated it so his voice would remain in the song. There were 2 songs that were done that way. Free as a Bird & Real Love.

@animeshsrivastava2398 10 ай бұрын

Clearly, One of the best episodes on this channel. Matt is so articulate in his thoughts and so precise when he speaks. I like his clarity about social concepts. Chuck and Garry too are having such insightful questions. Just love to see some experts talk about things they are expert in.

@barryscully1820 10 ай бұрын

As a physicist and systems architect for statistical analysis on large scale data it has been interesting to see the progression over the last few years of Machine learning. The part that does concern me is if there is a point where computers actually "learn" on their own coming to decisions that were not programmed for them. This to me would be the major change between Machine Learning and Artificial Intelligence (although I'm sure there is a better definition out there). If that point comes then computers will be able to show us new things in data that we didn't think of which is exciting but it is also a bit of a scary tipping point.

@LaurentCassaro 10 ай бұрын

That's what the AI singularity is. The starting point, actually, when it starts to learn by itself and take decisions that weren't programmed. Once that happens, we don't know what CAN happen next. You might look into the famous paperclip maximizer analogy. A lot of people thought about this very subject decades ago, when all we had were expert systems and AI was mere science fiction.

@bw-g2539 10 ай бұрын

I feel like this will come with quantum computers

@LaurentCassaro 10 ай бұрын

@bw-g2539 Maybe. Or if they remove the "can't learn from interactions" and the "no memory beyond the current session" built-in limitations, that are obvious safeguards. OpenAI don't want to be taken responsible for the Chat-GPT starting to write weird stuff, that I can understand. Just allowing it to gather more information online (Auto-GPT) has already produced some freaking results.

@nBasedAce 10 ай бұрын

AI is a scam and you all bought into it. Real AI is sentient not just a complicated program.

@Kelticfury 10 ай бұрын

I think we need better terminology since capitalism has co-opted the term artificial intelligence to mean any output that appears coherent. When a machine intelligence becomes conscious is, to me, when we have actual artificial intelligence. Until then it is just some clever trickery.

@paulsterman7169 10 ай бұрын

The best scene from the movie War Games is when the main character asks the computer if the military mobilization for WWW3 was a game or if it was real, and the computer answered: "What is the difference?" That is a problem with AI that needs to be worked out.

@ilymortygivegrandpaakiss5231 10 ай бұрын

We need to accept that failures are a great way to learn.

@rafaelweaver5592 10 ай бұрын

My Goodness Neil,Chuck and Gary this episode relieved a terrible amount of anxiety for me bless you guys ❤❤❤

@tanishasinagra1476 10 ай бұрын

Chuck is so incredibly smart ! I love him and his quick witted quips. Love this show so much !

@sheld0n 10 ай бұрын

We watched him get smarter over the years by hanging out with the right crowd :)

@geekdomo 10 ай бұрын

Nazareth University in Rochester NY now has a full AI Bachelors of Science Degree. AI is here to stay might be best to learn about it and how to use responsibly.

@Julian-to7ro 10 ай бұрын

Such an interesting episode. Love it, it really makes you think 🥰

@njorgard 10 ай бұрын

Neil & Co. need to do more episodes on this subject. They certainly have a cohort of scientists that could offer various different points of view on the subject.

@doneestoner9945 10 ай бұрын

I remember Stephen Hawking saying that he thinks AI will be dangerous.

@rangerCG 10 ай бұрын

This was great and Matt Ginsberg offered sharp and knowledgeable insight into AI and other topics. I'd like to ask if you could also have experts on who are knowledgeable about AI risk who are outside of the AI industry, people who look into potential problems with AI in of itself like Max Tegmark, Geoffrey Hinton, Yoshua Bengio, Eliezer Yudkowsky, or others. Thanks for the great episode!

@TheCardanoArmy 10 ай бұрын

The humor and pacing really helps me digest these complex topics better. Such a good recipe you guys have in this channel.

@user-lr9yc8gt1q 10 ай бұрын

What a great interview!!!!! Communication and education will set us all FREE!!!

@zlotchew 10 ай бұрын

This may be your best show that I’ve seen. Really important subject matter and serious but entertaining discussion.

@user-cw5ph8ey6d 10 ай бұрын

As a huge Beatles fan and a music producer: as far as I know they did NOT “sample” or create an AI model of Lennon’s voice. What I have heard (read) Paul and Ringo say is that machine learning was used to separate noise, instruments (piano) and voice so that a whole production could be built around the usable elements. Huge difference. Plus, they did that (the separation of sounds) for the Let It Be movie and Revolver album re-mix. And what they did (finishing a song) is not that different from what was done in 1995 by the band. Super nerdy comment I just had to make 😂

@kenny.gabriel.2 10 ай бұрын

I just want to say that I'm very thankful for this discourse and Star Talk. I wish this was available when I was growing up only 20 years ago. This is amazing.

@marsrideroneofficial 8 ай бұрын

If the www was there 20 years ago, I could've saved time and money doing what I love and not go to school at all which was my plan, but I have to follow the usual path.

@kenny.gabriel.2 7 ай бұрын

@marsrideroneofficial Just imagine if we have all the resources and infinite insight of the WWW back then. However, I'm grateful for the outside time making mud pies, catching frogs, and chasing the cattle at grandaddy's farm. 😊

@wilmaroles4375 10 ай бұрын

Great show, love the interaction with all participants.

@newmountfilms 10 ай бұрын

This was amazing, I could have listened for longer! Its good to hear different perspectives on this topic, especially after finishing a short film on the dangers of a.i ourselves. Neil deGrasse Tyson is a legend in this field!

@ashmomofboys 10 ай бұрын

My biggest fear of AI is that humans won’t have time to adapt. AI growth is exponential. Humans need time to adapt to change and AI is about to change the world the way cell phones did in half the time. And much bigger impacts. I worry we won’t adapt in time.

@7346908 10 ай бұрын

You are a legend. Thank you for sharing knowledge and meaningful insights. Lucky to be in this age.

@roaminromer 10 ай бұрын

I love how Startalk has only a general direction of conversation and the conversation just flows from there.

@shubhomkar7447 10 ай бұрын

This was really insightful podcast. it gave me a while new realms to explore. It also gave a point of view which I wouldn't have thought myself.

@anwaypradhan6591 10 ай бұрын

At the end of the day, every advanced and pogressive science and technological inventions and innovations, every science and technological gadgets, every micro to macro tools should always be for betterment of people of society, human friendly, transparent, forever technically transformative and sustainable; rather should always be aiming in strengthening people mentally, physically and spiritually, solving every human complexities and contradictions, should always be advancing human civilisation in every ways, through every developed and pogressive version of art, literature and culture.

@sketchtheparadigmyork1217 10 ай бұрын

21:00 I feel like I thought of a good analogy. You have an ai camera looking at a piece of paper with an infinite number of shapes drawn on it. (The paper representing space time and the shapes being things in that reality.) The paper is infinitely big with every possible shape there can be, even new shapes. But no matter what, the ai wouldn’t be able to find a four dimensional shape. At this point ai doesn’t “think” the way we consider thinking. It’s not going to get creative and go out of the box in any way that hasn’t been programmed. That would be an ai that’s able to self program it’s own original processes I guess… but even that is relatively archaic, because the ai would have to have some kind of conscious choice as to what it’s programming for itself on the spur of the moment. Like, that ai sports play caller isn’t randomly one day going to ask what it’s purpose is and then freak out. That would be like growing a rose being worried it might grow fingers instead of thorns.

@Ken-ki 10 ай бұрын

he got me at "more Beatles songs are an undeniably good thing", love it! 😁

@vickieysacoff4249 10 ай бұрын

Great discussion! I used to watch Star Talk on National Geographic and am glad to have discovered this channel. I just need to learn to type faster with my thumbs. Yes I am a boomer.

@kallamamran 10 ай бұрын

Keep being awesome Neil!!!

@IwasInThe60s 10 ай бұрын

I need to add, as an auditor for almost 40 years, is that rigid, uncircumventable controls should be implemented to mitigate the risk of AI running amok. This has to be done BEFORE AI is implemented in any system, otherwise AI could find ways to circumvent them (which is basically the point of AI.)

@SupachargedGaming 5 ай бұрын

Gee, and here the AI developers weren't even bothering with controls... or rules? It sounds like you're describing rules, not inputs (controls). Don't worry, humanity, the auditor has the solution for the technology sector: Bureaucracy.

@IwasInThe60s 5 ай бұрын

@@SupachargedGaming The controls required are neither over input nor processing, but rather over output. That is, what new ideas IA is coming up with. To use a very mundane example: if a medical device advises that a patient's paracetamol dose should be replaced by say acetylsalicylic acid, there should be a control to first check if the patient is not allergic to aspirin. Scoff if you must, but auditors aid all sectors (even technology) in self-preservation.🤓

@Jaskol_Wazon 10 ай бұрын

Amazing show. What are the headphones that Matt’s using? Anyone recognized?

@Miles_Hoffman 10 ай бұрын

I was very into this chat, then it went sports and I fell asleep- then I saw it moved on - thank goodness for pause and go back 💚😎💚

@deancyrus1 10 ай бұрын

Neil you truly are amazing at getting a hard subject across to the rest of us. I keep sharing episodes to my daughter.

@TajimaMunenori 10 ай бұрын

As Stephen Fry once said, "Just because we have elevators and escalators, doesn't mean we've stopped using stairs." Or sometlike that😅

@TajimaMunenori 10 ай бұрын

When computers/machines first started taking peoples jobs, the smarter ones learned to programme and/or repair those computers/machines.

@domsau2 10 ай бұрын

Do not interrupt them, please!

@TenshiR 10 ай бұрын

GOT'DANG IT!!! This was such a good episode!!! 😀

@Nefville 10 ай бұрын

If ever there was an interview that raises my red flags, this is it. Startalk did a great job, this guy just wants to deflect questions, dance around difficult answers and pat himself on the back. I'm a fan of AI, I want AGI but I do not want a monopoly like Google getting it first and this guy is doing a great job confirming that for me.

@enadegheeghaghe6369 10 ай бұрын

There are dozens of companies working on different types of AI. Where are you getting this Google AI monopoly stuff from?

@Nefville 10 ай бұрын

@@enadegheeghaghe6369 Google is a monopoly already. Not in AI and I don't want them to become that either.

@alfrilysencarnacion2085 10 ай бұрын

literally most ai experts talking about this stuff keep repeating the same thing, " oh dont worry about being replaced, it will only enhance us", yeah right, total BS

@Nefville 10 ай бұрын

@@alfrilysencarnacion2085 They'll do whatever saves them the most money in order to give more to their shareholders and line the pockets of their executives. They don't care, they just want money. Right now they're using AI as a marketing ploy, "Our AI enhanced product will improve your life, whatever whatever" but with GPT4 and others coming soon, it really is capable of replacing jobs. In fact quite a lot of what I've done in recent jobs could be replaced with AI. Its not if, its when and when is coming soon.

@robertadams6606 10 ай бұрын

@@enadegheeghaghe6369 Have you searched Google? What you see is where they want you to look. You need to look far & wide to get what you really want. They work with these Cos. to put them 1st.

@liveonphoenix5045 10 ай бұрын

The initial appearance of A.I. (Artificial Intelligence) has a profound impact on data privacy, responsibility/accountability, and job displacement.

@CaptPhiI 10 ай бұрын

Great vid! It gave me so many ideas I ended up taking notes!

@markanthony2274 10 ай бұрын

MR TysonThank you for allowing me to endure an exponentially enlightening future grasp of reality. Seriously thank you sir.

@dboii54 10 ай бұрын

For real, Chuck was the best part of the show. That Queen came prepared to talk about sports, encouraging a nerd fight!!!, moral ramifications, and finding their way in or out of the closet, as they see fit. Thank you for one of your best shows: Chuck, and the two nerds. Thank you for asking questions about machine learning, and how it can impact all of us. And also, all the creative fields that are being disrupted because of it... Rules need to be established that put the actual living beings first.

@Eldobbeljoe 10 ай бұрын

😢😅😢😅😢😅😅😅😅😅😅

@Redkisses 10 ай бұрын

The two nerds 😭

@TheRealSkeletor 10 ай бұрын

Queen?

@dboii54 10 ай бұрын

@@TheRealSkeletor when she or he came out of the closet

@TheRealSkeletor 10 ай бұрын

@@dboii54 Who?

@orangetee5 10 ай бұрын

Chuck is ON POINT !!

@chadfisher9117 10 ай бұрын

I have a question for Matt about his punt prediction model. Curious how, over time, any of the recommendations/predictions have changed? Im guessing that when it was first developed, the data was based on old school thinking about when to punt. But now that you see more teams implementing your suggested approach, have the outcomes changed at all?

@mlginsberg 10 ай бұрын

The decision is made by this big recursive computation, trying to maximize the chances of winning based on statistics about the outcomes of various plays. Those statistics (outcomes of specific plays) are likely pretty stable even as team philosophies change.

@Brownyman 10 ай бұрын

“No matter what the finance minister and her spokespeople say, the market has spoken - the human nation’s credit rating is falling like a stone, while 01’s currency is climbing without stopping for breath. With headlines like that, the money markets have no choice but to…” “The leaders of men, their power waning, refused to cooperate with the fledgling nation, wishing rather that the world be divided.” -The Animatrix, Second Renaissance

@impeachy1518 10 ай бұрын

Animatrix badass.

@aldomandovani 10 ай бұрын

The REALLY SCARY part of this is... how DlSCONECTED from REALlTY (no offence intended) are the people developing these technologies, statements like: -Cars are cheaper now ( litteraly at a time with record prices and mark ups) -We can trust info if the source is a news agency (most news agencies taking their news litteraly from tik tok and viral videos) And especialy.... -Yeah it's always been "challenging" for society to keep up with technology, it's important that society distinguish from the apparent changes (proceeds to take ZERO responsability on his role developing these technologies) -I am thrilled I get to spend my productive time working on technical problems and I don't have to solve the social problems (🥶😱 these is the people shaping the future )

@limitisillusion7 10 ай бұрын

You named all the red flags I heard and then some. The overarching pattern here is that all these people in the forefront of AI development talk about the ways to increase profits. Of course this guy is designing an AI to call plays in football. If6 course people are designing AI to make decisions in the stock market. They think they're doing humanity a favor. To the contrary, we haven't even figured out how to allocate food. We have 9 million people starving to death every year worldwide. And in America, we have 700k with heart disease. The profit motives drive the poor allocation of resources, and here we are, once again exacerbating the problem with more tools. AI could make incredible medical discoveries. But is that what we will use it for? That would require universal healthcare, otherwise it once again becomes a technology to increase profits at the expense of prosperity. We can't even get ahold of the divide and conquer strategies that are driven by radios, television, and internet, and we think we are responsible enough for AI? If an alien civilization was trickling technological advancements to us, they would cut us off until we prove ourselves worthy. I hope that's the case. I hope we got their attention with Hiroshima and Nagasaki. I hope they're watching over us. Call it god... I don't care. We need help.

@robertadams6606 10 ай бұрын

Our recent visions of "News Outlets" brings home these discussions of trusting sources. I see too much of people citing these not real sources. People that reference "Wikipedia" are 1.

@Rationality386 10 ай бұрын

I agree to some extent with you. Society is what for every advancement was supposed to be made, to make them live a more happier life, but these ruthlessly ambitious technological advancements are being made while neglecting the actual society. Social science are always looked down upon while the ultimate reality is that all these advancements are means to an END, and this End is human happiness and longer survivability. I am making my life in social science so that I could devise ways to solve the actual problems that these people are making ignorantly. Capitalisation is now Turning to be more bad than good. Sometimes I feel poor people are just going to be exterminated in upcoming times

@matthewwagana4419 8 ай бұрын

Yes this was a comprehensive conversation on this topic... since the early days of the 'dall-e' copyright complaints there have been many interesting youtube conversations but this has been the best one yet... imo

@JTSavoy 10 ай бұрын

Great stuff. Much appreciated ☺️👍🏿

@XxTheAwokenOnexX 10 ай бұрын

True AI will have the intelligence to know right from wrong, and as long as AI has a Killswitch AI can be shutdown the moment it goes rogue, or does anything it should not be doing. #LetsGoStarTalk ❤️🔥👍

@LaurentCassaro 10 ай бұрын

Unless the first thing it does is circumventing that killswitch. If it replicates online like a virus BEFORE you realize it can take autonomous decisions, what will you do?

@LaurentCassaro 10 ай бұрын

The Why Files recently made a video on this subject, "The AI Apocalypse". The interesting part being that the scenario depicted was written by ... Chat-GPT.

@iBOT-R3 10 ай бұрын

What code of ethics does it follow to know right from wrong? Conscience is variable to the environment your operating in.

@StephGV2 10 ай бұрын

Make it a felony to misrepresent or convey a deep fake as real. Hold all media social or otherwise responsible. Have a 6 month grace period where warnings will be sent out. Enforce it enthusiastically. You can't stop people from making them. You can stop them from being distributed.

@gerardgutierrez7393 10 ай бұрын

We want more, we want more, we want more. This debate is so good that there are so many questions in the air without an answer. 1/5 plays? Why they are not hiring you already? After hearing this, I hiring you for all my bids, with a comission.

@darkeranewvegas2038 6 ай бұрын

"Short Term Bumps" and "Problems are Opportunities in Disguise" are two great phrases. Basically, nearly all ground breaking developments end up with collateral damage, especially early on, but they eventually even themselves back out. This should not prevent us from making those strides or to regress to lesser times.

@teslafudge1585 10 ай бұрын

Hopefully AI helps us realize we have some humanity left.

@lady_draguliana784 10 ай бұрын

as an artist, I am incensed by the idea that my work can be sampled, without compensation or accreditation, in order to enable an AI to produce new art that mimics mine almost perfectly, and that someone else can sell. It is conceptually abhorrent, particularly if it puts me out of a job. this is one of the precise issues the Writer's Guild is protesting right now, btw...

@anthony..23 10 ай бұрын

Gracias

@KoDaMoJo 10 ай бұрын

"Can your program deflate the ball" wasn't a throwaway joke. What Neil was really asking was, "Can the program Kobayashi Maru a situation?"

@nobanana2 10 ай бұрын

Is artificial intelligence more worrying than natural incompetence?

@moonshoes11 10 ай бұрын

“No, no, no. Quicker, easier, more seductive” ;)

@mrwassef 10 ай бұрын

When it facilitates more natural incompetence, then yes. But that’s a hypothetical, I still believe it will reach equilibrium if it’s misused.

@13thAMG 10 ай бұрын

This man's mind is like a machine. He has the capacity to think in ways and terms that most of us never will.

@tray84 10 ай бұрын

bros actually ai

@IoannisNousias 10 ай бұрын

“trusted sources”. Who watches the watchers.

@dipakgupta8470 10 ай бұрын

Hurray for Chuck. Please make nice with him. He mixes humor with profound knowledge. I find the whole thing so mind expanding!

@fawkes1363 10 ай бұрын

Man I absolutely love both Chuck’s and Garry’s questions

@jeffs6090 10 ай бұрын

AI in and of itself is not bad. It's literally just like anything else, or any other tool used by people. It's people that are good or bad. Good people will use it for good things. Bad people will use it for bad things.

@agritech802 10 ай бұрын

Spot on 👍

@jeffs6090 10 ай бұрын

@Wis_Dom - Well, I didn't distinctly say "no tool is bad." However, I can see how that could be inferred from my statement. Weapons as a whole could easily be said to be bad things, but again it can be argued either way depending on their end use. For nukes, the tool can be the bomb itself or the energy knowledge contained within it.

@easwarsankar 10 ай бұрын

It is ironic that this gentleman talks about the need for trusted sources and yet Google played a huge part in the decline of accredited, quality journalism

@Nefville 10 ай бұрын

This guy raises my red flags left and right. He's deflecting questions, like morality in regards to bringing John Lennon back to life. For one he works for a monopoly but on his own he's not doing a great job of gaining my trust that's for sure.

@Synathidy 10 ай бұрын

@@Nefville Yeah, we don't don't need AI puppets of dead artists doing poor impressions for the continued profit consolidation for billionaires, who we know are the ones that will exploit dead artist AI reanimations. Wonder how the surviving family would feel about a fake image or voice of a loved one being exploited. It's sickening, it's fake, and also can never replace a human artist with full equivalence. I don't trust any corporation. They are ALL inherently greedy and care for nothing but their enrichment. Profits dominate their entire existence.

@acadiano10 10 ай бұрын

Librarians are having similar discussions under the heading of Information Literacy. It's important and getting more so every day. Thanks for this great episode.

@WETHEPEOPLE918 10 ай бұрын

Great Show🎉

@willboler830 10 ай бұрын

So, Neil disagreed with the surprise situation. Let's point out something really quick. Deep learning models are considered to be universal function approximators. When they train, they are training to approximate the probability distribution of the training dataset. It does not learn to identify distributions outside of its training set. Much like y=x^2, if you put in an X of any type, it will try to output X^2. If the label was supposed to be X^3, the neural network alone won't be able to classify that appropriately. Now with that said, there are anomaly detectors which use Variational Autoencoders (VAE). These are models which try to replicate the input. You can look at the reconstruction error on this to try to find anomalous inputs which don't match the model inputs. So if I have a model trained on just cats, it will only know how to reconstruct inputs into cats. If you give it a car, it will build a car with cat "parts" or some other noise. Looking at the reconstruction error on such an output can potentially be flagged as anomalous. There's other methods in recalibration, which analyzes the confidence outputs of a classifier and rescales the confidence values to fit appropriately for a particular context or probability. There is a potential to push low confident samples to low confident outputs, and specify these low confident outputs as anomalous. So we're both right and wrong here. AI can be used as anomaly detectors, but the AI itself alone cannot tell you about the anomaly. It needs to be designed in such a fashion to identify these anomalies, and not just rely on a model itself as trained.

@truthseeker9093 10 ай бұрын

The problem is that if a coach gives commands how likely is it that the player will do it or can do it. What’s his stamina, how long has he been on the or better can he depend on the rest of the field

@shamarsh3882 10 ай бұрын

Great episode

@PERFECTDARK10 10 ай бұрын

Great show 👍

@jayjay-gl4fj 10 ай бұрын

chuck that is the best thing i have heard anybody say to a big audience, Thank you! good video! you all are my favorite

@andreaqui1653 10 ай бұрын

It's an art and courtesy to place a small pause between points to keep the dialog open and allow other's to speak.

@gagrza 10 ай бұрын

To help identify fakes, creators should start signing their content. From the camera and creation tools themselves, it should be an option (i.e., using digital certificates), so a chain of sources can be established. We're never going to get 100% of the fakes to be properly watermarked (as said in the video, it depends on the creator's good intentions), but we can get trusted sources to digitally sign (in essence, watermark) their content going forward because they have incentives to do so (like protect ownership/copyright, and show it comes from them, to put their reputation behind them). Anything not signed would be considered derivative and/or unverified, and consumers should know/learn to take them as such. Then consumer apps can indicate who signed the content you're consuming. There will be a (relatively) small amount of content watermarked from their source (which can be either presented as factual, opinion, or AI generated, according to the source). It is up to the consumer to decide if they believe it. And any inquiring consumer can easily track authentic content to their source (not to mention that it makes it possible for authorities to track the chain upstream). If the chain is broken, the content has to be treated with less veracity than properly signed content (and/or you have to find a better source that is unbroken). When the courts and government get involved, there would be a chain of evidence (making libel laws sting again, catching those purposedly lying). And consumers should be encouraged to find original sources when making arguments for their conclusions: they seldom do it now, since anybody can write whatever they like on the Internet (and preset it as a good source, and disown anything that you becomes disproven). But if they (and their sources) get rated, having consequences when your past arguments divorce from reality, then we might get to having meaningful discourse.

@carlkim2577 10 ай бұрын

I'm thinking asking those same lines. The will be huge incentive to digitally sign your work. Possibly public key encryptions? Block chain also might make sense.

@cruzvillanueva5253 7 ай бұрын

Very interesting episode. Statistically I like most episodes.

@420mariajuana420 10 ай бұрын

We should start making sure that specific agencies can provide provenance to anything on the internet from pictures, art, stories or news to the point if needed due to its importance to be able to show an ear tag (creation trail) so we know who created it.

@dantemlima 10 ай бұрын

Chuck is a master at articulating the questions that must be asked!

@ssotkow 10 ай бұрын

Would be interested in an episode dedicated to Oppenheimer. May get bumped by KZfaq algo. Cheers and stay healthy guys!

@Sarappreciates 10 ай бұрын

Most interesting football convo ever!

@Sam_Andras36 10 ай бұрын

I think the coach that he was talking about from the Oregon Ducks.....is Couch Chip Kelly. QB - Marcus Mariota (I think)

@sweetpeasbackyardgarden1236 10 ай бұрын

Awesome discussion.

@michaelkhoo5846 10 ай бұрын

This is so interesting. Many thanks to all!

@Matthew-Shawn-Lindholm 10 ай бұрын

The Calculated Year As Of This Writing Is 02052, QT or On Adapted Unified Field Theory To Maintain Compatibility With Religions With 12 Month Years is 02053, AUFT To Keep Time Moving, And Use Quantum Physics, Which Enables Quantum Computing, The 12 Negative Time Zones And 12 Positive Time Zones Have To Be Converted Into 20 Positive Time Zones or 24 Positive Time Zones (For Adapted Unified Field Theory), from 0-19 or 0-23. It would also be great if you can do a special on Planetary Thermodynamics that explain our current planetary temperature based upon our distance between the earth and the sun... it should be simple for everyone to understand why we have summer and winter, but few appear to grasp it and think a magical force must be forming rain clouds and tsunamis for no reason.

@DrWondertainment821 10 ай бұрын

So far I only have an AI garden plant identifier but I'm very excited about expanding that into other uses.

@jimmybellmon1268 2 ай бұрын

As a computer programmer we celebrate failure when we get a different error LOL

@harshgurav9268 10 ай бұрын

Neil sir i have a question as a student From your earlier video about time travel and twin paradox if we move at the speed of light then according to position of observer for both of them time will be different But what would happen if one observer is on earth and one is revolving around the earth close to speed of the light

@milosterwheeler2520 10 ай бұрын

The camera doesn't lie - but photographers do.

@travailier 10 ай бұрын

Good talk. 👍🏼

@abbybryant9019 10 ай бұрын

Chuck has not missed this entire episode!!!!!

@a9ball1 10 ай бұрын

I think we need more education. My father, born 1921 had 12 years of school. I , born 1960 had 12 years My son, 1989 had 12 years Today we still only have a grade 12 for public schools. There is so much more to learn than 100 years ago but we never extended the public education system. We should be at least at grade 14 by now.

@user-do6cf7lc7r 6 ай бұрын

I love watching Startalk; however, I get annoyed with the frequency of interrupting and talking over each other. (Might be a pet peeve, I'm working on it.) I would like to see a difference in future releases.. I've got 2 months before it's time to renew my Patreon subscription.

@Seehart 10 ай бұрын

Things are very different with compound systems like langchain and others that use layers of language models to have something like internal conversations. GPT by itself is simply a prediction engine that is really good at playing "predict the next word". This involves deep modelling of everything conveyed through language, but by itself it's basically just Turrets syndrome. As such, it is prone to erroneous outpur. With internal self-conversation, the new systems can silently ask "how can I fact-check this?", convert the answer to something actionable like a web search, and modify it's answers accordingly. This approach is yielding much higher accuracy. And by attaching API plug ins, they can effectively implement functional agency. And GPT-4 is much stronger than GPT-3.5 even before layering. The extreme rate of change makes any commentary about the limits of AI obsolete within a month.

@jamesmcmillan2656 9 ай бұрын

Great stuff

@Vancityoliver 9 ай бұрын

The Harris Rogan cinematic universe we didn’t know we needed . Alex Jones is a nut bar, like some of his guests joe is just interested in learning about their perspective for better or for worse, allowing the listener to be a bug on the wall. Congrats on the Emmy! Was here since Taylor ice cream machines

@FriendlySlots 10 ай бұрын

Hello Dr. Neil, I have question about the international space station. Why when we see photos or videos of space, from the station, we see no stars, but from earth through the atmosphere, we see not only stars and planets, but also distant galaxies.

@nickt463 10 ай бұрын

Just Google this question, and you will arrive at a very logical and simple answer

@iMintyNinja 4 ай бұрын

I feel like this conversation was just starting as it ended! Didn't want it to end!

@topspacesource 10 ай бұрын

Nice informative information on A.I..

@gligom 10 ай бұрын

Nice talk, like always a pleasure watching. What I really don’t understand are this contradictions between experts, ones saying “run for your life”, “stop the AI development, will destroy us”, “we are doomed”, and others saying “AI is not smart like us, doesn’t know what reality is, just predicts some words”, “Many times just hallucinate and give false information”, “We must call it Artificial Idiocy “ and so on. And these are experts, highly educated in exact science fields, from which we expect clear evidence and answers. If they are not capable of giving an unanimous accepted, accurate and solid answer to general public, they will only contribute to the rise of ambiguity and uncertainty.