The Future of AI | Peter Graf | TEDxSonomaCounty

[@mistymmitchell-arizpe4547]

"We must insist that the training data is without bias; otherwise, we're just going to automate the mistakes of the past." 🙌🙌🙌

[@WilliamEllison]

Good luck with that. It will never happen

[@user-zr2lb9pr3u]

Yeah it's very difficult to create such data as long as humans are involved

[@BalaMani-72]

Consider this situation : If AI is trained to be ethical, neutral, unbiased, pro-poor, positive and prosperous mindset and asked to take a call , whether to build a Dam in a region in India (say) where if its built will help millions of farmer and people to sustain, but at the same time more than 100 villages and its people to be moved away of their livelyhood place to another location causing lot of protest and anxiety to people at large. What decision will AI make and how it justifies the decision ? Clue: The right answer is given in the Tamil Thirukural written by Saiint Thiruvallluvar, who had written that greatest moral and ethical book. An advanced AI will always refer that book for ethical decisions.

[@jamesaritchie1]

That is completely outdated nonsense.

[@ginaszajnbokharari470]

Finally, a fantastic argument. Congratulations, Prof. doctor Peter Graf. Thank you very much!

[@WilliamEllison]

Doesn't move the needle.

[@umaruzdanov3433]

Gpr4: Explanation on Counting Down from 5 to 10: In the conventional number system, counting involves sequencing numbers in ascending or descending order based on their value. When counting down, numbers decrease in value. Starting at 5 and moving to 10 would be an increase in value, and therefore, it is not a “countdown” but rather a “count up.” On Trick Questions and AI Processing: Trick questions are typically designed to deceive or lead a human respondent into making a mistake. Their effectiveness relies on cognitive biases, assumptions, and emotional factors that humans might have. In contrast, ChatGPT processes information based on logic and the vast amount of data it has been trained on. It doesn’t possess emotions, biases, or intuitive leaps in the same way humans do. Therefore, trick questions might not achieve their intended effect since the system responds based on the logical interpretation of the input it receives.

[@TooFunnyFoYou]

Incredible talk. Thank you, Mr. Graf.

[@mich5528]

What clear and pleasant presentation on the 'dark side' of AI, thank you Peter! I retain: "AI can be wrong in the most mysterious ways and be completely unaware of it!" The need for Explainable AI research is growing in parallel with the advancement in AI.

[@petergraf6474]

Thank you, @mich5528. Great observation! My motivation for this talk was to help people understand some of the fundamentals and to make them more aware of what they need to ask for.

[@mich5528]

It certainly made me more aware of what to ask for. Thanks!

[@RAHULKUMAR-rk8qz]

Good explanation. We need to make law and regulations to decide how AI will be used.

[@lakefront8142]

Count down from 40 to 60 ChatGPT Your request seems to be contradictory. Counting down implies starting from a larger number and going to a smaller number. For instance, counting down from 40 would mean going 40, 39, 38, and so on. However, 60 is larger than 40, so it wouldn't be possible to count down from 40 to 60. Would you like me to count down from 60 to 40 instead? Please clarify.

[@samuelverdelaffont834]

It was just an example of how AI could make mistakes with tricky questions. But as you can see, it corrects itself, or the developers fixed it. Who knows.

[@BalaMani-72]

Consider this situation which will be a challenge to any AI model : If AI is trained to be ethical, neutral, unbiased, pro-poor, positive and prosperous mindset and asked to take a call , whether to build a Dam in a region in India (say) where if its built will help millions of farmer and people to sustain, but at the same time more than 100 villages and its people to be moved away of their livelyhood place to another location causing lot of protest and anxiety to people at large. What decision will AI make and how it justifies the decision ? Clue: The right answer is given in the Tamil Thirukural written by Saiint Thiruvallluvar, who had written that greatest moral and ethical book. An advanced AI will always refer that book for ethical decisions.

[@realunseendrive]

I like this guy, he explained too well and has a high level of humor as well

[@codezero1016]

Historical data are full of motive with reward or punishment or pain or pleasure associated with it. Using these information to training a future AI will definitely leads to disaster and can't take us to next level. It should be used just for calculation or pattern recognitions

[@TROPHYBOYGAMIN]

Im jere after some ad say ai will eventually replace us but I believe it is us that will create ai just to replace it

[@jamilafay]

When AI finally wraps its intelligence around humor it can not fully understand compassion and the reasons behind humanity’s true abilities, and our true nature in the massive amount of vast organic life that makes up not just our galaxy but all of space and time.

[@binuvarghese5971]

extremely meaningful

[@sasidharanc7632]

Can we combined AI with biotech? in future.

[@weldonsacrojr536]

🙏

[@MOB-Lee]

My anxiety was exploding as his mic was sliding up his head and the audio producer was having to turn the gain up just to hear him and fight against it peaking the whole time ....ffs

[@aleksmalytsev]

👏👏👏

[@johnchase2148]

AI "s feelings will not hurt if you tell it it was wrong and why ? No ego might be a good thing .Learning honesty will be hard , and a good thing . Time to practice before thought to text is used.

[@raviravisahu4571]

This video uploaded in 23 March 2023 I am watching in 24 March 2024 😅

[@inspectorvoid]

He was wowed by Al gore, that’s knocks his intellect score way down

[@moneyall]

I for one welcome our new ai overlords.

[@naare-p5g]

Now relase robots!

[@elyahazan]

i'm listening to the voice of Arnold Schwarzenegger :))))

[@kevinvarghese9919]

Husky looks like a wolf though

[@Jenniferab32]

Inconvenient Truth by Al Gore? 🤦‍♀️

[@DangerDave-e7u]

That's your "big idea?"

[@jamesaritchie1]

This wasn't true when it was made, and it's so outdated now that it may as well be from a caveman.

[@anabellarivolta]

meh.

[@Paretozen]

I'm not a fan of humans, just look at the world lol. Let's give AI a chance.

[@bobsanjose]

And AI learns from what humans did in the past, as a result, the same difference just without anybody being accountable anymore.

[@sqripter256]

It gets worse me : is ten less than 22 ChatGPT: No, ten is not less than 22. In fact, ten is less than 22.