Home Creators Posts Import Register Favorites Logout

Content

Apologies for the post hiatus but we're coming back swinging. Leave any questions you have about this video in the comments and we'll answer it in the aftershow!

Files

why AI can't pass this test

Thanks to Fiverr for sponsoring this video! If you are curious about AI and are looking for a place to bring your project to life, head to https://fvrr.co/answerinprogress to check out the services available to you and use the code ANSWERINPROGRESS for 10% off! How intelligent is artificial intelligence? In this video, Sabrina answers that question and explores how AI (large language models like OpenAI's chatGPT and Google bard) understand and solve problems. She figures out where AI is smarter than humanity and where it isn't, and calls her shot on whether artificial general intelligence (AGI) is around the corner. SUPPORT US ON PATREON https://patreon.com/answerinprogress SUBSCRIBE TO OUR MONTHLY NEWSLETTER https://www.answerinprogress.com/newsletter SOCIAL MEDIA Sabrina Twitter: https://twitter.com/nerdyandquirky Instagram: http://instagram.com/nerdyandquirky Melissa Twitter: https://twitter.com/mehlizfern Instagram: http://instagram.com/mehlizfern Taha Twitter: https://twitter.com/khanstopme Instagram: http://instagram.com/khanstopme CREDITS Produced by Sabrina Cruz Research Assistance by Tom Lum (https://www.tomlum.com/) Video Editing by Joe Trickey Motion Design by Sabrina Cruz Sound Design by Joe Trickey Special Thanks to - Colt Kirwan - Evie Yannakidis - Gregor Reynolds - Kariza Santos - Melissa Fernandes - Toby Hendy (https://youtube.com/@tibees) Fiver Freelancers I mentioned in this video - Ahmed: https://www.fiverr.com/abdeldzns - Sasha: https://www.fiverr.com/conversionai - Thomas: https://www.fiverr.com/tomwalchak_nlp - James Brown: https://www.fiverr.com/britishvoice4u MUSIC Epidemic Sound. Get started today using our affiliate link. http://share.epidemicsound.com/answerinprogress RECOMMENDED READING Common Sense, the Turing Test, and the Quest for Real AI (2017) by Hector Levesque On the Measure of Intelligence (2019) by François Chollet TIMESTAMPS 00:00 why sabrina is beefing with AI 00:29 why AI has gotten smarter in the past few years 01:10 oops it was capitalism 01:47 what is intelligence 02:22 how we're measuring intelligence 03:49 a reflection on my capability 04:21 asking for help with Fiverr (our sponsor) 06:32 trying to use a printer in 2023 06:52 writing 5 exams and answers 600 questions 07:29 freedom 08:00 we are all shackled to the chains of content 08:18 humbling myself with some friends 09:01 results day 11:54 wouldn't be an AIP video without a research montage 11:59 sabrina realizes this is more work than expected (shocked pikachu) 12:41 what makes an exam difficult 13:24 why AI might find some exams easy 13:50 why AI might fail some exams 14:30 why we can't just keep adding training data (long tail problem) 15:16 how can we make AI smarter 17:55 why AI isn't smarter than us, for now 18:16 thanks again ------------------------------------------------------------------------------------ Welcome to the joke under the fold! Here's one signal for the machine uprising What happens when an AI takes a picture of itself? It becomes selfie aware. Leave a comment with the word AWARE to let me know you were here ;-)

Comments

Juan

Hey guys, loved the video as I usually love all the AI related content Sabrina does. My question here is related to being sentient, so I've always understood AI to be closer to a robot that regurgitates knowledge more than an entity that actually learns and understands, this video seems to back that idea, but do you personally think that sentient AI is, eventually far in the future, a possibility?

Lyndsey L. Werner

This might just be more evidence that tests are not very useful evidence of knowledge. When I heard “Tom” I thought you meant Tom from Let’s Learn Everything. 😅 Then I got confused because Fiverr had a Thomas who you worked with. Then on a rewatch I saw it DID say Tom Lum! (In my defense… I mostly do other things while I listen…) Do you ever try to “break” the AI? By which I mean, do you ever stress test it to see what causes it to “hallucinate” or see if you can force it to? (Completely unrelated, but I’ve been wondering for a while now. Can we get the soup dating sim or a backlog of the news letter? There’s a couple things that were in the newsletter I’d like to see. But the exclusives require a password I don’t have? 😅 ) (I had some bad typos that changed my meaning a lot! Like missing the word “not”!)

Sam Parsons

Unrelated to AI, but I’ve lately been inspired by you guys as well as some other amazing content creators to start making my own video content. Is there anything you wish you’d known when starting out, or anything you’d rather not have learned the hard way?

Blaze Utz

Short question: where can I find that shapes and squares test? I'm curious how I would do on it

Mitchell Bell

Any questions in any of your tests that gave you a "Why am I doing this" moment.