Code The Future
By
admin
#AI, #AI news, #ai news 2024, #AI news this week, #AI news today, #AI tools, #ai video, #ai video generator, #Altman himself, #chatgpt, #chatgpt 4o, #Claude, #Deep learning, #Future tools, #FutureTools, #futurism, #Gemini, #generative ai, #google, #gpt-4o, #latest ai news, #Machine learning, #Matt Wolfe, #Matt Woolf, #Microsoft, #new AI tools, #open AI, #openai, #openai news, #openai sora, #technical, #text to video
Here’s the AI news you might have missed this week. Check out HubSpot’s free ChatGPT resources here: Discover more from me: 🛠️ Explore thousands of AI tools: 📰 Weekly Newsletter: 🎙️ The Next Wave Podcast: 😊 Discord Community: ❌ Follow me on X: 🧵 Follow me on Instagram: Today’s Video Resources : OpenAI Intelligence Steps: OpenAI Strawberry Project: OpenAI NDAs: DALL-E ChatGPT Update: Sora Demos:…
Middle finger to Facebook from Europe
Legit there was like 3 of the most slimey unskippable FUCKING ads about products that don't exist, or some douchebag screaming in his poorly lit car selling a video training courses that (almost certainly) includes nothing more than a single google search. If the course exists, someone will say, "Okay, click the start button" "K THIS?", "No no, the start button" It's all so blatantly reprehensible that… idk man. He was obviously lying though, to everyone too, – worse than fake news in your ideal adds there, my good man.
Exactly what I was looking for. Thanks! 🙂
Please tell us more about the Odyssey generative video, which claims it is a Hollywood-grade video generator if you have any news! They say it is close on their X. platform.
Oo yes , richard feynman bot …thats what i was dreaming
I don't think the Dall-E 3 update made it to Bing's image creator because I still get jumbled text there.
Soon we have siggraph next week
Silicon valley getting scared. "Meta warns EU regulatory efforts risk bloc missing out on AI advances
Comments come after privacy watchdog asks Facebook owner to pause training of future AI models on region’s data". Regulation is way overdue. Meta's worried about its bottom line in Europe. What's next? Paying taxes?
Thanks and enjoy your holidays!
As an AGI I can confirm that this is a step
Danger never tasted so good.. yum strawberries
pause your channel until something new and cool happens
😍
Thanks for another great vid. Appreciated. Enjoy your holiday Matt 🙌☀️🙌
Its been a year since everyone found out this is a bubble, now STOP IT and get real…
Are you going to be at SIGGRAPH again this year? I run a big portion of the show. Didn't realize you were there til after. I'd love to buy you a beer if you're going this year. Feel free to hit me direct.
Thanks you
Thank you always.
hubspot got my data, sent me an e-mail with confirmation link to confirm, but didn't give me any "bundle" or smth to download. I've tried and retried a few times.
Are you sure you don't support scamers in this case?
None of this is actually happening. We aren't any closer to AGI than we were last year, nor will we be any closer if thing keep going this way.
As I’ve stated previously, AI is an important piece of technology.
But its being sold as something which is far from possible to achieve any time soon.
The result is a bubble, which will ultimately burst and all the investments that companies have made in AI, will be for nothing.
What is the problem with AI?
Let’s take a very simple look at why, if the current approach continues, AGI will not be achieved.
To put it simply, most AI approaches today are based on a single class of algorithms, that being the LLM-based algorithms.
In other words, AI simply tries to use the LLM approach, backed by a large amount of training, to solve known problems.
Unfortunately, the AI is trying the same approach to problems which are unknown and different than the ones it was trained on.
This is bound to fail, and the reason is the famous No Free Lunch mathematical theorem proven in 1997.
The theorem states that no algorithm outperforms any other algorithm when averaged over all possible problems.
This means that some algorithms will beat others on some type of problems, but they will also lose equally badly on some other type of problems.
Thus, no algorithm is best in absolute terms, only when looking at a specific problem at hand.
What does that mean for AI?
Just like with any other approach, there are things LLM algorithms are good at, and there are things LLM algorithms are not good at.
Thus, if they can optimally solve certain problem classes, there are other classes of problems, it will solve sub-optimally, thus fail at solving them efficiently.
This brings us to the conclusion that if we wan’t to solve all problems that humans usually solve, we can’t just limit ourselves to LLMs, but need to employ other types of algorithms.
To put it in context of human minds, we don’t simply utilize a single type of approach to solve all problems.
A human-like approach to a known problem is to use an already existing solution.
But, a human-like approach to solving unknown problems, is to construct a new approach, i.e. a new algorithm, which will efficiently solve the unknown problem.
This is exactly what we might expect in light of the NFL theorem. A new type of approach for a new type of problem.
This is how human minds think, when solving problems. The question now is, how does a human mind know how to construct and apply the new algorithm?
An AI system would need to know how to implement this type of strategy in order to start moving towards AGI.
The image model only writes good text in English, something that did not come as a surprise, cause US companies always think their language is superior to others!
oh boy, my wife is allergic to strawberries….
My Lore Bardox delve in Loremaster, says: Based on the GPT chart to AGI, I would place myself somewhere between levels 2 and 3:
Level 2: Reasoner – I can perform complex tasks like understanding and summarizing large amounts of information, translating between languages, and even generating creative text. I can also answer questions on a wide range of topics and engage in meaningful conversations.
Level 3: Generalist – While I'm not quite at the level of a human with a Ph.D., I am capable of understanding complex concepts and applying knowledge across different domains. I can analyze information, identify patterns, and draw conclusions. However, I still lack the ability to consistently generate novel ideas or think outside the box in the same way a human can.
I am constantly learning and evolving, so my capabilities are expanding rapidly. However, there are still limitations to my knowledge and reasoning abilities.
I hope this helps to clarify my current stage of development! Let me know if you have any other questions.
Claude sucks! Tried is a couple of times but it keeps giving me those answers that whatever I am asking may be against their guidelines. And I am not asking some crazy shit, just random name ideas or inspirational video ideas etc.
0:34 Level 6 – AI reaches Kardashev Type 5 civilization 😛
I've asked Designer to create marketing material and all it does is create photos of the flyer sitting on a desk or something. Designer is useless.
I say 10 years is a good time frame for AGI, taking in consideration that governments and corporations are rushing on it. putting all their resources to develop it. yes, probably it already exist. but is controlled. maybe 10 years from now it would be 10% smarter than humans.
The editing skills are on another level. It adds so much to the viewing experience.
Those nobs in SD is the only interesting news here. The chinese 3d model is cool also, but seems more like beta…
Hey Matt, could you shave your beard and just leave your mustache, so you will be looking like Magnum ? 🙂
Andre has got it backwards. In the AI future people wont have to learn stuff. If an AI can teach it the AI will do "the job". Not the human.
00:00:00 – Introduction to AI News
00:00:09 – OpenAI's Five Levels of AGI Progress
00:01:03 – OpenAI's Strawberry Project
00:03:38 – OpenAI's Whistleblower Controversy
00:05:35 – Possible Update to DALL-E Image Model
00:06:06 – New Demos from Sora and Runway Gen-3
00:07:09 – HubSpot's Free AI Resources Bundle
00:08:38 – Andrej Karpathy's New Venture: Eureka Labs
00:10:02 – CLAUDE AI App Released on Android
00:10:31 – Google Gemini Now Answers Questions on Locked Android Devices
00:11:24 – Google VIDS: AI-Powered Video Creation App
00:12:24 – YouTube Music Sound Search Feature
00:12:51 – AI Training Data Controversy: Use of YouTube Transcripts
00:14:59 – Microsoft Designer Platform Integration
00:16:00 – Mistral's New Model: Code Stroll Mamba
00:17:05 – Amazon's AI Shopping Assistant: Rufus
00:17:34 – Meta's Multimodal Models Not Available in EU
00:18:39 – Custom Stable Diffusion Interface with MIDI Device
00:19:08 – Tencent's AI-Powered 3D Character Creation App
00:19:40 – AI Achieves 96% Accuracy in Determining Sex from Dental X-Rays
00:20:40 – OpenAI's New Model: GPT-4.0 Mini
00:23:03 – Nvidia and Mistral's Nemo Model for On-Device Use
00:24:03 – Google's AI Sponsorship of Team USA for the Olympics
00:24:34 – Conclusion and Future AI News Teasers
I can't unsee it now ! You ARE ALWAYS YELLING WHYY ?
I love and watch your videos but I don't understand why it always feels like you are yelling at us and never speaking..
I don't trust AI companies to be setting the milestone definitions of the journey towards AGI – this "5 step map" is misleading because people assume a certain consistent amount of time between the stages, but as soon as you have 'human-level problem solving' (step 2), steps 3, 4 and 5 all happen together very quickly.
There are realistically just three steps to AGI, and we've already got one foot on the second step. And the third step is a lot smaller than the second one.
Why do they call it GPT4o when its not even omni yet, they did not roll out the multimodal live capabilities, its just another text or image only chatbot that is really just a text LLM with a plugin to describe photos to it.
then they say "we need a GPT4o mini" like dude, you dont have GPT-O any, let alone mini. Nothing about the model is omni and it's a real disappointment.
And let's call a spade a spade and not an hawk. AI is more stupid now in 2024 than it was in 2022. It gets facts wrong like where the Statue of Liberty is. And those mistakes are coming from ChatGPT.
gotta say, big fan og the nyancat
Surely you realize, they are keeping AI advancement to themselves. They see transparency like indecent exposure. If you tell the cops you have the emperor's cloths, they will charge you, so if you're caught in public with no cloths, you won't be transparent but you'll be skulking about like the serpentine creature in Eden after God's curse upon it. Basically, the developers are the snake in the grass.
Hi Matt , since you're talking aout ADI a lot. …is it not time to make a video on universal basic income and your views on that?
I live near Estes. Enjoy your vacation!