Musk: Shocking AGI PREDICTION! & Is SHADOW MODE Dead?
ฝัง
- เผยแพร่เมื่อ 28 มี.ค. 2024
- Get Magical AI for free and save 7 hours every week: getmagical.com/drknowitall!
In a recent interview, Elon Musk predicts AGI smarter than the smartest human by NEXT YEAR, and Super Intelligence smarter than all of us put together by the end of the decade! Plus is Tesla's Shadow Mode no longer active? The evidence points to yes, and it's actually a big deal not just for Tesla's Full Self Driving (FSD) but for their Optimus Teslabot as well!
**If you are looking to purchase a new Tesla Car, Solar roof, Solar tiles or PowerWall, just click this link to get up to $500 off! www.tesla.com/referral/john11286. Thank you!
Join this channel to get access to perks:
/ @drknowitallknows
**To become part of our Patreon team, help support the channel, and get awesome perks, check out our Patreon site here: / drknowitallknows . Thanks for your support!
Get The Elon Musk Mission (I've got two chapters in it) here:
Paperback: amzn.to/3TQXV9g
Kindle: amzn.to/3U7f7Hr!
**Want some awesome Dr. Know-it-all merch, including the AI STUDENT DRIVER Bumper Sticker? Check out our awesome Merch store: drknowitall.itemorder.com/sale
For a limited time, use the code "Knows2021" to get 20% off your entire order!
**Check out Artimatic: www.artimatic.io
**You can help support this channel with one click! We have an Amazon Affiliate link in several countries. If you click the link for your country, anything you buy from Amazon in the next several hours gives us a small commission, and costs you nothing. Thank you!
* USA: amzn.to/39n5mPH
* Germany: amzn.to/2XbdxJi
* United Kingdom: amzn.to/3hGlzTR
* France: amzn.to/2KRAwXh
* Spain: amzn.to/3hJYYFV
**What do we use to shoot our videos?
-Sony alpha a7 III: amzn.to/3czV2XJ
--and lens: amzn.to/3aujOqE
-Feelworld portable field monitor: amzn.to/38yf2ah
-Neewer compact desk tripod: amzn.to/3l8yrUk
-Glidegear teleprompter: amzn.to/3rJeFkP
-Neewer dimmable LED lights: amzn.to/3qAg3oF
-Rode Wireless Go II Lavalier microphones: amzn.to/3eC9jUZ
-Rode NT USB+ Studio Microphone: amzn.to/3U65Q3w
-Focusrite Scarlette 2i2 audio interface: amzn.to/3l8vqDu
-Studio soundproofing tiles: amzn.to/3rFUtQU
-Sony MDR-7506 Professional Headphones: amzn.to/2OoDdBd
-Apple M1 Max Studio: amzn.to/3GfxPYY
-Apple M1 MacBook Pro: amzn.to/3wPYV1D
-Docking Station for MacBook: amzn.to/3yIhc1S
-Philips Brilliance 4K Docking Monitor: amzn.to/3xwSKAb
-Sabrent 8TB SSD drive: amzn.to/3rhSxQM
-DJI Mavic Mini Drone: amzn.to/2OnHCEw
-GoPro Hero 9 Black action camera: amzn.to/3vgVMrH
-GoPro Max 360 camera: amzn.to/3nORGYk
-Tesla phone mount: amzn.to/3U92fl9
-Suction car mount for camera: amzn.to/3tcUfRK
-Extender Rod for car mount camera: amzn.to/3wHQXsw
**Here are a few products we've found really fun and/or useful:
-NeoCharge Dryer/EV charger splitter: amzn.to/39UcKWx
-Lift pucks for your Tesla: amzn.to/3vJF3iB
-Emergency tire fill and repair kit: amzn.to/3vMkL8d
-CO2 Monitor: amzn.to/3PsQRh2
-Camping mattress for your Tesla model S/3/X/Y: amzn.to/3m7ffef
**Music by Zenlee. Check out his amazing music on instagram -@zenlee_music
or TH-cam - / @zenlee_music
Tesla Stock: TSLA
**EVANNEX
Check out the Evannex web site: evannex.com/
If you use my discount code, KnowsEVs, you get $10 off any order over $100!
**For business inquiries, please email me here: DrKnowItAllKnows@gmail.com
Twitter: / drknowitall16
Also on Twitter: @Tesla_UnPR: / tesla_un
Instagram: @drknowitallknows
**Want some outdoorsy videos? Check out Whole Nuts and Donuts: / @wholenutsanddonuts5741 - วิทยาศาสตร์และเทคโนโลยี
Get Magical AI for free and save 7 hours every week: getmagical.com/drknowitall
I would be very surprised if Tesla wont use shadow mode for training on edge cases.
All reasons you mentioned why they won't use it, also supplies to training on interventions or disengagements. Bad drivers can disengage the system wrongly, as much as shadow mode decisions won't match drivers outputs... In both cases you will pick data from good drivers.
I think shadow mode is still here, and contribute a great deal in training supervised FSD ...
Five years ago I thought my chances of seeing AGI and beyond were slim to nil. Now I am excited that I will witness this new era in humanity even though I am in my late sixties.
Yeah agreed I am 62 and was expecting it in the 2040s which would be at the end of my life, however its looking more likely next 2 to 8 years from now
Let's hope what you witness isn't SkyNet coming to life!
@@vanodne Yes the possible downsides are also terrifying, even getting to a good outcome is going to be bloody difficult, job disruption alone will cause massive social rifts.
It could be the best thing to happen to humanity or the worse. I'm hopeful it's the best. These next few years or less! Wow.
@@vanodnewe signed our death sentence already. I am incredibly sad knowing what world my little niece will be part of
I expect that Shadow Mode will continue to operate in the shadows, looking for more and more edge cases, just as it has been doing.
If AI is thousands of times more powerful than humans in 5 years and it’s anything like Google Gemini, we’re in trouble…
Emergent behaviour could be where it really bites us, having capabilities that we can't predict
What we as humans also need is an AGI in the Medical domain maybe MedAI. Restricted to the medical practice for research purposes.
AI has modeled every known protein, this is a huge leap forward. Now is a matter of simulations of protein interactions to build working molecules. COVID vaccines were created this method.
You are right. Better medical advice and treatments and lower costs
Ashok mentioned shadow mode in the v12 Elon livestream back in August
he absolutely did, came here to say this myself. 👍🏻
People outside of the Tesla/AI bubble chamber have no idea what is coming. We are seeing changes coming that are going to utterly transform our world and for the vast majority of people, this is going to be terrifying. Faced with job loss, or rapid job changes, many are going to do a knee-jerk reaction to neo-Luddism and there will pandering politicians who will cater to that reaction. I do not see any politician that is ready to face this problem. Mishandling this transition is going to be tragic. We have the opportunity to enter a period of incredible abundance but it is going to require wisdom and steadiness in our leadership and I do not see any of them capable to handle this. Most are not even discussing it. Consider that even if a hundred million people are aware of the coming impact of intelligent machines, there are 8 billion people out there, that is only about one percent of the population.
I find it a bit ironic that we are spending billions and billions to build AGI, and billions more to maintain it. All in order to match, and then replace humans - which are far more energy efficient (better for our planet) and are core to our capitalist system (they spend the money that they earn). It makes you think - what is the end game here? If AGI replaces 50% of human jobs.... switching millions from earning money to requiring government support.... will our economy shrink dramatically? will tax revenue shrink dramatically? how will society support the unemployed?
I think Rbt Kennedy who is running as an Independent might be our best choice. At least Elon seemed to entertain his thoughts about the needs America now has. I don't see anyone else I would vote for at least at this point. Plus he's more aware and agrees what Elon is doing.
@@andrewvh6809AI powered bots will produce far more & cost far less than human based systems. Means abundance & massive reduction in many prices due to increased capacity & competition. Don’t need as much money & still have far better standard of living. Will take time
Great summary. As Elon states the books of Iain M Banks are brilliant. Player of Games...or..Look to Windward perhaps give best insights into a very favourable AI human relationship
Also google 'Notes on the Culture' by Iain M Banks..10 min read maybe
Shadow mode or no, they must have a method of analyzing discrepancies (interventions) to get better in edge cases.
7:39: "Oh great AGI ...", is there a god? ....... compute compute compute ...... "There is now."
@DrKnowitallKnows - Hi John, your mic volume seems to be a little low on the past few videos. It's particularly noticeable when you have a guest or share a clip. As always, I enjoy your show. Thanks for the efforts!
His mic is almost always too low. Listen to the difference between his voice level, and the outro music. The music isalmost always booming by comparison
Dr Know-it-all, thanks for taking the time to make this video and share your knowledge with us about Ai and how it is evolving. Really informative. i’ll be sure and check out your sponsor... Magical Ai👍
Jon, I have been told LLM models are passing the bar exam. I know that isn’t the highest bar, but.
No need for shadow mode when you collect enough data from interventions/disengagements from FSD driving itself. Early on there were not enough FSD drivers and it was so far away from driving itself that it was useful.
I think I shadow mode is still running in campaign mode when your manual driving but only saving data based on a campaign request
It is quite possible that FSD may 'log differences' like shadowmode did, so shadowmode may be a built in feature of fsd, just not a primary task, but a secondary task.
Your logic makes sense, Dr K-i-a! I bought FSD in 2019 and had been disappointed ever since because it would make critical and egregious mistakes. However, 12.3 is a whole different beast. It is so much better and I am excited about it again. For the first time I actually see a path to level4 and beyond. You might be right, it might be behaving better than humans in most instances. It might also be possible for Tesla to replace shadow mode learning with just the verbal feedback that beta users provide after an intervention. I think with LLMs they can now curate that feedback and quickly figure out frequents errors and work on them.
Yeah makes sense. If so where does fsd get its LR input now? This question has nagged me for some time.
Doc ... 2 things are not being mentioned.
1 hw4 is not being mentioned but hw3 still being used. I believe hw 4 is being reworked
2 the tesla training computer is being replaced by Nvidia chips
This might be the first time I've seen Elon underestimate time scales negatively (imo), the rate of hard to imagine breakthroughs is going to be off the charts, never before seen speed.
I can’t believe they got rid of Springfield mall
Intelligence measured as the ability to process a single task is not a sufficient measure. AI will not only process a single task faster and better than humans, in ten years the AGI will have real time access to all the data of humankind and be able to process that in real time as a single integrated intelligence. It will be well aware that no human or group of humans is qualified to manage its function.
It won't even be 10 more years. Millions of American jobs are on the verge of disappearing .
I think shadow mode is still used, because when you're driving you can see the predicted path of fsd (so why using this much compute for only graphical reasons?) and you can always catch for dangerous or obvious divergence. If the human slam on the break but fsd no, it's dangerous and obvious. If the human accelerate a bit quicker than v12, it's not a big deal depending on the focus of the training. And false negative can be caught by human to improve their "divergence finder". You can't directly use disengagement data, their is either a processing needed (to reconstruct the good behavior) or a collect of human driving in similar situations (but it's better if you pick the data in which humans diverge from fsd in these similar scenarios)
No need for shadow mode. Tesla switched to log interventions instead. These can be annotated & are more useful
This is level 4 already. Level 5 will be soon. Maybe 1 to 2 years or even sooner since it's learning exponentially
Phenomenal 🦾👀
Isn't shadow mode essential for gathering clips from the fleet to train the system? E.g. if you want to gather clips of cars coming to a full stop at a stop sign, you need a piece of software that you can tell to look for those situations. It seems to me that FSD itself is perfect for that. It is already very good at detecting stop signs. And stopping for stop signs is a relatively easy problem, but you also want to query for more intricate traffic situations. Running FSD in shadow mode seems the most proficient way of detecting those situations.
They could still use shadow mode to demonstrate where FSD would have prevented an accident
Loved that little comment about Apple because it's true. They are really falling behind, aren't they
We already have some great solutions for big problems like climate change but it’s politics and the extremists that get in the way. Maybe when AGI says the same thing people will hear it better coming from a politically neutral source.
Check out Delton Chen. He’s got a way around the political gridlock.
They'll just call it "woke" or claim that it's being controlled by elites.
One way to get more human intelligence would be to educate the people who are already alive to be more intelligent. Generally, we work at a fraction of our capabilities. And AI could help us become smarter.
Will help us become smarter.
Optimii will make incredible teachers.
Human teachers will need to learn to code, or something. 😂
Hahahaha 🤭 ” WHO the hell knows what apple is doing .. I don’t believe they are doing anything “ 😂😂😂😂😂😂 now that was funny I could not stop laughing
They can also train from video even video they create so they can just create training videos that can show just exactly what they want.
Shadow mode. Maybe that's why they're doing the download the software and everybody must have a demo drive because he wants the software downloaded on everybody's car right whether they buy self-driving or use it or don't. So that way they can run Shadow mode on any car. I bet if the software isn't downloaded they can't run Shadow mode
If aI is improving 10x every 6 months. That means it’s improving 100% every 19 days. Is that right? WOW
I know they had a ton of data labelers and that kind of went away (at least I think). I wonder how they curate good quality data from hundreds of millions of miles they get all the time? I suppose telemetry might be a start (e.g., hard breaking). Is there an algorithm that looks for pausing or slowness?
OK, I should have listened to the end of the video. ;^)
"looking at a very different world in a few years" I'm concerned that none of our politicians (worldwide) seems to have a 'plan B' for that scenario. They don't even talk about it ...
I suspect they are using shadow mode in all the vehicles without FSD. This would allow them to assess the quality of the human driving in those vehicles and collect training video. I think, given how FSD is performing now that using shadow mode FSD 12 to curate fleet data for training would be very effective. Need to improve U-turns? Set a trigger for all vehicles where FSD is inactive to send back clips of humans doing U-turns within a certain range of shadow mode. Otherwise, there is a lot of FSD hardware being wasted.
Agree 100% and I think this is why they’ve avoided pricing FSD so “everyone” could have it.
Cars without FSD collect data about how to drive well, and FSD users intervening collect data about perceived mistakes FSD is making. 👍🏻
Is AI compute and AI IQ the same, or directly proportionalitet? Never heard about such a relationship. Elon just talked about compute, right? I would guess that it is Möre exponential, or maybe, unfortunately, an S-curve.
Ray Kurzweil predicted this 20 years ago. But thought it wouldn’t be till 2045
Elon may or may not be the smartest but he sure is the most productive.
Boy isn't that the truth. There has been very few people in the world who have been as productive as Elon and we probably will not see another person like him in our lifetime.
Elon is a genius of intellect and execution.
A wicked combination!
The Lex Friedman podcast with Yann LeCun have some good point why real AMI, is not just around the corner, LLM can't be an AMI/AGI no matter the level of compute, compute != intelligence, we will see a lot of improve in AI, but it will not be an AMI/AGI
I think when it embodies a human form and interacts with the world AGI will happen. Don't you or do you have another thought?
🙏🏽
What is shocking here? Come on.
I dont think we will have AGI in 2 years. I dont even think we define AGI correctly as it is a glossy explanation of smarts that does not cover everything. For example, the creativity of people is not well covered. Yes some forms of AI can use random generators or extensions of a trend to get what looks like creativity but nothing like General Relativity could be produced in any forseeable upgrade path. And speaking of Turings' definition of AI, that is too simple. Talking to the AI and not knowing it is AI or people is not that hard in some cases and not that easy in others (Ask the AI to tell a story about some childhood adventure, etc).
Then there is the problem of making it too human like. We want it to be accurate but then we tell it not to do SOME things people do (prejudice, lying, cheating, exaggerating, etc). We really dont want to add emotions to the AI do we? They cause us so many problems. And many times the solution to a problem is just not what we want to hear. All these things are intertwined with AGI
Dont get me wrong, the current AI is amazing and far better than I would have predicted a few years ago. But I dont see how LLMs can become really intelligent. They just predict the next best word to say based on reading a big data set. It is an amalgam of the data, not smarts. It is more of a synthesis of the data on that subject.
Here’s a question: Can an AGI be smarter than all humans if all of those humans are wrong about something? That is, all experts agree on a topic, but they are wrong. What would the AGI train on that would drive it to be correct?
It wouldn't "train" the same way neural nets do now. It would be able to reason as we do and come to its own conclusions. Information will still be a limiting factor for the AGI, and lack of certain info may be the reason experts are wrong. However, the AGI may be astute enough to figure out the information gaps.
Exactly. Human knowledge built on examination of physical world, on experiments, data collection. AGI could easily be better at interpretation of data.
Next stage to theorise gaps needing research & carry out such.
Will AGI have artificial ethical constraints on research? Or Grok style freedom?
@@TheSpartan3669 I think Super AI will be friendly and the period most concerning to me is the time AI moves from human intelligence until it goes beyond human intelligence. That will be a period when we will not have the capability to stop it from being destructive if that's what it chooses. With the help of Neuralink maybe we can stay at par so that doesn't happen.
I was wondering this for a long time, AI is trained on human knowledge.
1. How does AI the knowledge trained on is good or bad?
2. What if all data is used to train on, how does AI train further? Humans can experiment, and learn from the results. AI can do that, but knows the result only from simulation, or it has to be embodied and experiment like humans do..
Sub question, who have seen the movie _"Transcendence"_
(With Johny Depp in 2014)?
So good !
Digitial intelligence is about solving deterministic or statistical problems.
There is no love, no hate, no compassion no morality, no humanity in AI.
Thus we need to understand AIs limits and gaurd against its misuse and abuse by humans in every part of the world. Unfortunatey we will need to use AI to combat these misuses as well as appropriate laws and regulation of the production, distribution and use.
Also the same for its training and inference.
Otherwise we are all screwed.
That's as far as we know and understand now. Things could change so rapidly with self learning we really don't know if or how emotions will develop in AI.
@@sandybayes well it will be a very inhuman form of emotion but I assume if it been given a voice then its emotion can be vocally communicated. It it is in a humanoid form then it could be given facial motions etc or even teer ducts to cry, but would it be genuine emotion since it could never feel phyical and or psychological pain.
It really does push us to really think and quiery were these things happen in humans. Our physiological responces having an abstraction in our brains/minds and remapped into a equivalent emotion "pain" and indeed when humans suffure huge emotional /psychological pain then this being transformed into physical pain.
The cli-fi novel (tome) Ministry for the Future has AI solve the climate problem. As in any good novel there’s lots of drama so the answer isn’t introduced until Chapter 42 😂. The AI comes up with a psychological twist that uses human nature (greed?) and monetary policy (money creation). It’s called Carrots and Sticks. Turns out people don’t like taxes but LOVE rewards! Whodda thot!
AI training on human knowledge will think Harry Potter is more important than the Bhagavad Gita because more copies of HP sold last year. It will be a mess and unused.
the age of hyper abondunce? That is the scary part. I have so many discussions with people that are in the 40s and 50s. Even younger some time and their definination of live is their work. Most of them can´t a find a thing to do when they have abundance of time or money. That is realy hard for most people. I would do a lot of things if i had time and money or some kind of income that would make me not dependend on work alone. But for most people that is the scariest thing. WHAT WOULD I DO. And this is said. And ultimatly this will stop this revoulution in some way. Throught politics throught anger and so much more. And it will not be the technology fault, it will be ours to adapt to that change.
The technoloogy will be the easy part. the rewirering of the human self will be the hardest.
We already know for the most part the solutions to climate change, but are unwilling to implement the necessary changes on a global level. Having an AGI that we can ask for solutions won't help much if we are not willing to implemented the suggested solutions to this global problem.
Hopefully Neuralink will be used to improve human intelligence so there is a meeting of minds with Super AI to improve the world we humans have worked hard to destroy. I'm very hopeful with Elon continuing at the helm.
The real question will not be whether AGI comes up with solutions to, for example, climate change but whether humans will want to follow the suggestions AGI makes.🤣
You need to do an episode of why Tesla has not released a "WAZE" killer yet. Or why Tesla hasn't released a "Tesla maps" app yet. Not only is this a huge potential upgrade to Apple or Google maps, it would also be a good revenue stream.
My guess is that Google made Tesla sign a non-compete clause in order to get access to Google Maps. Surely it isn't open ended, but has a sunset clause built in. FSD was released in 2020. Therefore, a 5 year noncompete would make sense. So 2025 before Tesla can release one. Could be a 7 or 10 year noncompete, depends on how much Tesla paid Google.
Yes, Tesla is gathering second by second road details from 5 million cars. They already are years ahead of Google maps etc. I googled my house and the photo shown was 2 years ago.
Tesla created maps, assembled from multiple passes of a spot from different cars, in different directions, at different times of day, as demonstrated at each AI days as part of their simulation system, is a large part of making full autonomy work.
Imagine local level knowledge of the entire roadway of the planet, updates nearly realtime.
Route around detours, know when intersections are common busy in advance, potholes, etc.
Google uses the data of every phone with Google software (location on), its own mapping, and traffic sensors. It was estimated that well over a billion users kept their location on.
Also, I'm sure that many Tesla drivers wouldn't be comfortable having their car cameras used constantly to map areas.
@@TheSpartan3669 Google doesn’t have 5M+ cars with all around cameras collecting road data.
Tesla maps will be far better than Google maps in time.
I read and agreed with the vision put forth by Ray Kurzweil in 2005. The reasons were irrefutable and what's happening now was inevitable. I believe consciousness is behind everything and we are purpose driven to ultimately achieve a common goal. All unknowns can be known and data contained in computised matter, all the matter in the universe actually, allowing the God Head to awake go into the mother of all black holes and seed a formless new universe with what is known so far.. Sex life of a universe
Elon time
can you segment the video on the bar at bottom of screen so we can skip the ads?
Tongethar my husband ❤
When AGI becomes so intelligent not sure I want Einstein doing the house chores for me. What place will humans have in a world where there are a zillion overqualified Einsteins running things. We will feel pretty stupid .. how will humans avoid feeling useless!!? 😮
I still believe that one build of FSD will never be safe enough. FSD tuned for California will be different than FSD tuned for Texas will be different than FSD tuned for NYC will be different than FSD tuned for Paris will be different .......
Until Tesla announces FSD ver 13 or FSD ver 14 is adopting this strategy, we will continue to see complaints of wildly variable results across the USA, much less the globe.
And, because of Chinese data restrictions, we already know that the Chinese version of FSD "WILL BE DIFFERENT" than the US version, by definition. The Chinese government has prohibited Tesla from exporting camera data to the US for training, thus requiring FSD for China to be trained on China only data.
I'm pretty sure they are already doing this. Some users reported that FSD drives very differently in New York than in smaller cities.
Another clue is that FSD uses a different stack on highways and the switchover is seamless.
Part of the input is is GPS coordinates, which means FSD will always drive like a local.
Elon has said many times that you will be able to drop a Tesla anywhere in the world and it will drive.
I don't think it works that way because it all depends on compute right now but may change in time. A human can make the changes needed for a change in locale, why not Super AI?
@@andrasbiro3007 Correct.
Bradford Ferguson financial said this on his cross country trip.
There’s zero reason to believe it wouldn’t operate this way.
Once you see it’s not a technical issue, in fact it’s the ONLY way this could possibly work.
I think the “it’ll never work” crowd don’t understand AI’s mastery of dimensionality, and/or the scenario doubling capacity each bit of input adds.
8 cameras mean the inference input has somewhere around 8 megapixels at, I believe, 10 bits per pixel RGB or 30 bits per pixel, so 240M bits of input.
Adding just 64 bits of latitude/longitude gives you sub 5M radius training precision, plenty of accuracy to allow for specialize learning from challenging and unique local conditions, i.e. avoiding the “you’re not from around here” phenomenon.
Now add an additional few hundred bits for time (including years), temperature, precipitation, altitude, etc. and you have a system that can drive the world’s roads with near perfect local knowledge that no human could ever achieve.
I’ve heard people say that the same system that drives in California could never drive in India.
They were close to the truth but nevertheless missed it by a mile, because they’ve never understood that AI is incredibly good at learning and inferencing in multiple dimensions. 🫡
That’s way to long of a commercial bro
Thumbs down - find it repulsive that you do an advertisement on your channel - don’t we have to see enough mind numbing ads everywhere.
This is all good as long as we have electricity available and no EMF attack. If we lose power. We are basically back in the stone age but with a lot of wiring.
A faraday cage will stop EMP, standard requirements for military installs for past 30 years.
Thank you for pointing out that horrific shit may happen in the future.
Never considered what nuclear war could do the world before you wrote this poignant response. 😂
How would you know you could ever trust such an Oracle? If the Oracle knew that the solution to a given problem could only be achieved by the elimination of much of humanity, would it tell us that? I think it would be much more likely to tell us that the problem can be solved by doing X then Y, etc. All the time knowing that one of these steps would lead to an extinction event. Would we be smart enough to see through its subterfuge?
BeWare of Oracles!
More fear for country dictators around the world. N Korea, Russia, Israel, Venezuela, China, India getting closer.
Wow, nobody has ever made that argument before, thanks for enlightening us all!
Failure of AI alignment is literally an existential threat because an artificial superintelligence that decides to pursue its own goals and deems humanity a hindrance could easily end us. It could bio-engineer a lethal airborne virus or create trillions of nanobots that selectively target humans or simply guide humanity to its own destruction over millenia.
The intelligence gap between us and the AGI could be larger than the difference between us and ants. And why would a super AI take orders from ants? Ensuring alignment of the AIs goals to ours is critical, and we may only get one shot.
@@tommornini2470 You are the King of subtlety and I appreciate that humor.
I do have a pentient for stating the entirely obvious like it's a new thought.
😝
@@nickmcconnell1291 As I appreciate your ability to subtly mock my response. 👊
The smartest person on earth is Edward Witten. This is nearly universally known.
Deep thought, anyone…anyone
But Elon has consistently made very inaccurate predictions right ? His 5 might be 10-15 if it’s similar to his predictions abt FSD.
I thought that Elon Musk recording was a fake. Din't sound like how Elon talks at all.
And SpaceX were suppossed to be landing man on the moon this year...
Take Elon's guesstimates with a pinch of salt
The next bull run is coming for EV companies. NIO will be TSLA's main competitor and the 2nd leading car manufacturer in the world by 2035. Once TSLA releases the Compact TSLA in late 2025 this will cause demand to soar. Everyone in the world will be driving fully autonomous electric cars by 2040 produced primarily by TSLA and NIO. The stocks have both been crushed by demand concerns, but a large scale compact affordable EV hasn't even been developed yet...we are only at the beginning. The rapid transformation is coming.
you're all hallucinating too much 😂
Get Magical AI for free and save 7 hours every week
The key for humanity will be how we use AGI, not how capable it is. If we ask the right question(s), like how to solve climate change -- and then act diligently on the answer -- it will become a society-advancing tool. If we use it for war, it will become a society-ending weapon. The choice is ours... for now.
We don't know AI for that. We know what to do, what we lack is the will to act.
The key is not how we use AGI. The key is what data is used to train AGI.
We ARE using AI for war already and I’m sure AGI will be as well. Fortunately, Human Caused GW is a human created illusion to manipulate the scientifically ignorant which is the vast majority. So war or bioweapons or government created famine will end most.
Climate change😂
@@andrasbiro3007AI may provide answers that are less disruptive than the means we already know how to do.
Elon has been wrong about the timing of FSD for so many times I would take his prediction with a grain of salt.
You’ve clearly not experienced v12.
Tesla turns impossible into merely late.
@@tommornini2470FSD doesn't even exist. And what has Tesla done that was considered impossible?
Annoying kid in back seat - are we there yet? You promised me beach day, where is it? You lied.
AGI might have more compute but it doesn't translate to being more intelligent because AI has no agency. Also. If compute 10x every 6 months it means it will increase 1 Billion fold over the next 4.5 years. That is highly unlikely to continue growing at that pace. A trend that cannot be sustained won't be sustained.
People said that about Moore’s Law since its inception in the early 70s.
While your statement is true in the limit, judging where the limit lies is extremely difficult.
So, it’s already smarter than a chimp or a dog. I would argue it’s as smart as an idiot savant because it understands language and can answer questions. I also don’t think it’s that power intense, as it can run on a laptop computer. It’s power intensive to train it, but that is starting at zero. How much energy did it take for humans to evolve from zero? We evolved in parallel, so it would be the sum energy of all humans that ever lived.
It isn't. Not even close. Dogs are able to self-learn, think abstractly, generalize, and solve novel problems. They have a good understanding of the world and leverage long-term memory to preserve and build that understanding. Above all, they have high consciousness
Neural nets are excellent at pattern recognition, but they aren't able to generalize existing "knowledge" and apply it to problems outside the training data.
🧩🎤
Another flat earth level of intelligence video
GPT-4 is already smarter than most humans. Way smarter actually.
Define "smart"? Currently AI uses the collective internets pool of knowledge to spit out answers to any questjon...such as Google. So if you say AI will be smarter than any human in the future is incorrect...AI which is the same is today's collective internet has been smarter than any human for many many years if you define smart as knowledge