Will AI be THE END of us?

Sources used:
- Open letter: futureoflife.org/open-letter/...
- CBS full interview with Geoffrey Hinton: • Full interview: "Godfa...
- 2022's Expert survey on AI: aiimpacts.org/2022-expert-sur...
- Standford's AI Index report: aiindex.stanford.edu/wp-conte...
- Yann LeCun's proposal for new architecture: drive.google.com/file/d/1BU5b...
- Wait But Why blog on AI: waitbutwhy.com/2015/01/artifi...
- 📚 Superintelligence by Nick Bostrom: amzn.to/41JCjiM
- The AI Dilemma (highly recommended!): • The A.I. Dilemma - Mar...
- The Alignment Problem from a Deep Learning Perspective: arxiv.org/pdf/2209.00626.pdf
- Predictability and Surprise in Large Generative Models: arxiv.org/pdf/2202.07785.pdf
- Max Tegmark lecture on Life 3.0: • Max Tegmark lecture on...
- What should we learn from past AI forecasts: www.openphilanthropy.org/rese...
- ABC News interview with Sam Altman: abcnews.go.com/Technology/ope...
- Chris Olah on what the hell is going on inside neural networks: 80000hours.org/podcast/episod...
- X-Risk Analysis for AI Research: arxiv.org/pdf/2206.05862.pdf
- The Case for Halting AI Development | Max Tegmark on Lex Fridman Podcast: • Max Tegmark: The Case ...
- Sparks of Artificial General Intelligence: Early experiments with GPT-4: arxiv.org/pdf/2303.12712.pdf
- Goal Misgeneralisation: shorturl.at/ijmqx
Honestly, I feel a bit restless these days by the recent AI developments. It is very exciting and terrifying at the same time. Hence, I dove into this topic and did some research, with the hope to understand a bit better what's going on. AI risks are sometimes so conceptual and hard to put into words. So I hope this video will give you a glimpse of the discussions around these difficult topics and what still needs to be done to help make AI really benefit humans, instead of doing harm.
Hope you enjoyed the vid! 🤗
🔑 TIMESTAMPS
================================
0:00 - Open letter & Debate about AI risks
3:26 - When will AGI exist?
7:45 - AI forecasts have been wrong before
10:02 - Ways AI could be dangerous
15:37 - Solving AI alignment problem
👩🏻‍💻 COURSES & RESOURCES
================================
📖 Google Advanced Data Analytics Certificate 👉 imp.i384100.net/anK9zZ
📖 Google Data Analytics Certificate 👉 imp.i384100.net/15v9y6
📖 Learn SQL Basics for Data Science Specialization 👉 imp.i384100.net/AovPnJ
📖 Excel Skills for Business 👉 coursera.pxf.io/doPaoy
📖 Machine Learning Specialization 👉 imp.i384100.net/RyjykN
📖 Data Visualization with Tableau Specialization 👉imp.i384100.net/n15XWR
📖 Deep Learning Specialization 👉 imp.i384100.net/zavBA0
📖 Mathematics for Machine Learning and Data Science Specialization 👉 imp.i384100.net/LXK0gj
📖 Applied Data Science with Python 👉 imp.i384100.net/gbxOqv
🙋🏻‍♀️ LET'S CONNECT!
================================
🤓 Join my Discord server: / discord
📩 Newsletter: thu-vu.ck.page/profile
✍ Medium: / vuthihienthu.ueb
🔗 All links: linktr.ee/thuvuanalytics
As a member of the Amazon and Coursera Affiliate Programs, I earn a commission from qualifying purchases on the links above. By using the links you help support this channel at no cost for you.
#artificialintelligence #datascience #ThuVu

[@ryanwammons]

Incredible compilation, very engaging. Will be using this for my AI/ML students (Northern Virginia Community College--I'm the data analytics coordinator and currently sole professor of ML). Thank you for your contributions to the field!

[@Thuvu5]

Thank you Ryan!! So glad this material is ideal is useful for you!

[@robstewart8531]

Outstanding overview of the issues.

[@Thuvu5]

Thank you Rob! I’m glad you liked it!

[@MarcosGabrielSantosRocha]

Thank you to get us up-to-date and less anxious about what's happening with the advance of AI technologies! Watching your videos from Brazil 🇧🇷♥️

[@LukeBarousse]

Such a critical discussion to have on the impact of AI. 🙌🏼 We are all enjoying the benefits of this powerful tool RN (including myself), but we need to think about how we can control these systems for the good of society. Thanks for taking the first step with this convo, Thu!

[@Thuvu5]

Indeed! There are great tools that can be used for both good and bad things. Thanks for checking out the video Luke! 💛

[@MirMohammedAli24689]

I just found your videos and I like them!

[@GaryBarclay]

Succinct, comprehensive, efficient and elucidating.

[@dino_rider7758]

Very well put together video!

[@valor36az]

What an extraordinary review !

[@sergentumba583]

You are just amazing at explaining.

[@Anon-xz6hu]

Video quality is improving 🌚

[@josephvanname3377]

The key to AI safety is to first become a respectable human and learn about reversible computation.

[@MichealAngeloArts]

Great compilation! I'd like to have a similar style compilation on the Computer Vision AI advances if possible (e.g. Stable Diffusion, Control Net, Dreambooth, Midjourney, JAX Diffusers etc). This time however without the piano background (it's a bit distracting to me 🥲). All the best! 🙂

[@Thuvu5]

Oh I’m sorry for that! I’ll keep it in mind next time 😀

[@IM-rk9nm]

Hi Thu Vu, may I ask you.. are you directly talking to the camera as you go or are you using some sort of a prompter? Just curious :)

[@Thuvu5]

Hey, it depends on the content. For tutorial videos I usually talk to camera, while for more difficult, research-heavy videos I use teleprompter to get the facts right 🙂

[@Alvaro-og4rk]

Awesome video

[@maxmp7827]

Hi Thu, Nice video. I would like You making a Q&A video about AI.

[@jasonmulligan7059]

brilliant

[@cerealport2726]

It's a longstanding trope in sci-fi, that computers will kill us all. one of the first sci-fi books I ever read 40+ years ago, was by Stanislaw Lem (if i recall correctly), and was based on the premise of someone asking a computer to "help make humanity better". in the end, the computer decided humanity was best improved by turning them all into computers. I dont know exactly when he wrote it, 1950s or 60s, I think. I do think we need to approach this carefully. When I think what trouble I've caused myself by writing code that did something unexpected, and I magnify this to the level of an AI that has control over something, then... yeah, we're doomed!!

[@Achrononmaster]

Good comment, except the "doomed!" part. Doomed how exactly? Just pull the plug out at the wall (and don't watch too many scifi movies, they're distorting your realism). Also, we'll get rid of nuclear arsenals, just to be on the safer side. This is a benefit of advanced AI. What nuke capable governments under public pressure will not eventually scrap all the nukes? - for this very reason! It's gonna happen, because the AI genie is out of the bottle. (Shoulda' happened a while ago, since the humans with the control switches are even more twitchy than the AI. But hay, I'll take the fictitious AI excuse as the patsy here.) But also consider the massive opportunity cost of failing to push advanced AI. You cannot credibly use such extreme scifi to determine scifact. Meanwhile, it is a very real thing that advanced AI can automate laborious work, freeing up humans to do better things, and not-for-profit things. How? Governments are monopoly currency issuers, so cannot run out of their own currency, so can _always and without question_ hire anyone not employed in the private sector. To do what? To do anything we care to imagine in the public interest: there are infinitely many good things to do: permaculture gardening, community care, public arts, ... it's limitless, every day we wake up to a severe labour shortage - provided you expand the whole meaning of "work" - "work" can be non-profit, decent living wages for public purpose service. (Say hell-to-the-no to the highly regressive UBI in other words. UBI = a non-reciprocated claim on someone else's labour, usually a poor worker in the global south. Otherwise known as the neoliberal rentier wet-dream.)

[@cerealport2726]

Is the ability to pass exams really a sign of its power though? I feel like its a nice showpiece, but doesn't really demonstrate too much in the way of practical value. Perhaps I just haven't looked at the context under which these exams were passed, and am missing something important? For example, I passed an exam to be able to learn to drive a car, but I had basically zero skills to safely drive a car.

[@Thuvu5]

Yes I agree! I think the exams are more a benchmark for how well the AI can reason and make inference in various topics. But no guarantees they would be able to safely handle a problem in reality

[@Spartacus547]

it'll exist when programmers realize that the internet is not a mind it's just a library, it's not enough for a program to grab information, it has to contemplate the information "what is this?' "why" "who am I" and capable of coming up with new ideas and inventions not just referencing and plagiarizing anything on the internet I think that's a long way off, it's has know it self as "I have a self"

[@Thuvu5]

Interesting thoughts! Thanks for sharing 😊 I wonder, though, if it’s strictly necessary for AI to have sense of self to be able to effectively process information. But indeed I agree that being able to contemplate, or maybe ‘iterate’ some sort of thought processes, would make AI much better than it is right now

[@Pradonis7]

As data lovers, we have nothing similar to compare with, so, what will happen? In my opinion, We still need to wait to get more data to be able to try to predict it. Us or IA? 🤔

[@shorgoth]

The finger issue has been solved a few weeks after it appeared.

[@Eswar.]

Data analyst vs data scientist Which is best for freshers

[@Lie404]

Does AGI can develop emotion and conscious of being like a real thing

[@rrrevered]

Thank Thu Vu [data analytics], once again ❣ I know who will be right... 😉

[@otto7979]

Listen close to her description of the potential ways AGI could go wrong in a toxic way...don't humans with expanded resource pooling already pose those threats?...just manifesting in a slower timeline?

[@vuminh211]

wait just how many books have u read? you cite different books throughout different videos chị à

[@HoaNguyen-ji5fz]

❤

[@JohnVKaravitis]

What people fail to realize is that I telligence exists to solve problems. The stupidest person can refuse to cooperate with the smartest person, thus invalidating the smart person's plans. Also, any truly intelligent entity will realize that cooperation is more fruitful than betrayal.

[@snpjun5153]

ultron is comming

[@Leto2ndAtreides]

The Coast Runners example is definitely something that humans would also do if they can... Because it creates an easy advantage. It's not different from humans trying to cheat in school... Humans also just want rewards. The alignment problem is everywhere. Even so, that AI would go monstrously out of control seems unlikely... It may make mistakes - but other agents (and humans) will likely react to it. Plus, AI isn't actually going to advance to monstrous power in an instant... In part because there is no real path to gaining such power from human controlled systems in a short term way. Complex AI will evolve with the needs of the environment around them - as humans themselves have.

[@Leto2ndAtreides]

It would be more honest to say that the goals humans want from an AI are "obedience, subservience, and being a highly competent and attentive servant... That has no desires or ability to produce outcomes that aren't aligned with that of its master". "Honesty" isn't necessarily a thing that most people want from other humans or AI. Humans only want truth they like, not truth they don't like... Because being able to handle actual truth requires maturity that few humans have.

[@jason-the-great52]

Yet gpt-4 cannot even make a simple checklist app for android...

[@GungaLaGunga]

Keep calm. As long as they don't teach AI to code, connect it to the Internet, allow AI's to use API's, or do anything really stupid like that, everything will be just fine.

[@Thuvu5]

I’m afraid we’ve done all of that already 😅

[@GungaLaGunga]

@@Thuvu5 Yep, and oops. 🙃

[@Thuvu5]

@@GungaLaGunga haha yeah! exciting time..

[@amt.7rambo670]

Missing long hair thu vu.

[@rizvy99]

MY MAXIMUM WEB CONTENT IS FULL OF AI CREATION EVEN IN MY COUNTRY BANGLADESH ALL ELECTRONIC MEDIA BROADCASTS ARTIFICIAL PROGRAMMES WITH AI TO FIX MY MIND.