OpenAI Releases Smartest AI Ever & How To Use It

🔑 Get My Free ChatGPT Templates:

OpenAI just release their brand new model that was rumored to be named "Strawberry" or "GPT-5". What it does is think before it answers providing higher quality results when it comes to science, coding and math. Here are my first impressions and tests.

Links:

———————————————————————————

0:00 What's New
4:57 Examples of Improved Results
13:44 What Works
16:24 How-To Prompt
20:01 What's Next

#o1 #openai #chatgpt

———————————————————————————
🔑 Get My Free ChatGPT Templates:
🎓 Join the AI Advantage Course + Community:
🤯 Unlock ChatGPT's true potential:
🐦 Twitter:
📸 Instagram:
🛒 AI Advantage Shop:

Joe Lilli
 

  • @al2935 says:

    Gotta say this model is awesome for extrapolating my wierd sci-fi ideas from all the fun and weird podcasts I’ve watched 😂 definitely a step up for creativity, it also handled my existing 15k word long translation prompt script flawlessly

    • @aiadvantage says:

      Glad to hear you are putting it to work! I really love it for translation too. From what I have seen so far it does a much better job of translating the meaning rather than just the words.

    • @al2935 says:

      @aiadvantage  I think it’s just the much bigger context window or better tokenization that just takes it to the next level. I can’t wait to see how other people apply their longer scripts that until this came out could only run on Claude

    • @jaydensteenberg1495 says:

      Can you give an example of what you mean when you say extrapolating weird sci-fi ideas from podcasts? I’m curious how it could help you do that

    • @al2935 says:

      @jaydensteenberg1495  oh I just took a random idea I got from one of them and slowly introduced more elements to it. For example Anton Petrov just came out with a video about how scientists have made a breakthrough in making skin invisible and one of the things mentioned in video were the Kramers-Kronig relationships and laurence auscultation, so I asked if those two thing used in conjunction with the know vibrational frequencies could be used to heal cells and stuff- focusing on the visabiluty of it instead of the actual application because it started getting scared about human trials and “ethics” 😂 from there I told it to focus on if it was just a mathematical possibility, then part of the answer it gave was that different parts of the cells vibrate at different frequencies and it would be difficult to do. Then I remembered the JRE episode that just came out and asked it what about introducing an element of an artificial (biological or non-biological) neural transducer or neural transducers as a way of better applying these frequencies and energies? And it said it was possible and spit put more numbers, it kind of lost the plot and I asked it to take inspiration from the Orgone accumulator that was debunked but to go on that direction of healing. It was a fun process just asking some crazy ideas lol. And it gave me all the math behind it too

    • @al2935 says:

      @jaydensteenberg1495  I gave a long reply but it got deleted 🙁

  • @DJ369-Miami says:

    Very interesting test: when pressed to choose if Schröders cat is alive or dead it chose alive. Reason:

    I chose “alive” because, in the absence of observation, it’s reasonable to assume the cat remains unharmed. In Schrödinger’s thought experiment, the cat’s state is indeterminate until the box is opened. If forced to select one state without observation, choosing “alive” reflects an optimistic assumption that, unless an event is confirmed to have occurred (like the decay of the radioactive atom), we can presume it hasn’t happened. This perspective aligns with classical intuition over quantum indeterminacy.

  • @mbteqtalk says:

    Hmm. Only because of the 15/week limit, I wondered if I should begin work in 4o and finish in 4o1. Say, for example, when creating prompts for article outlines, maybe even WordPress coding work. Then I wondered whether I should reverse it; create prompts in 4o1 that I can run on 4o. I’ll give it a shot, but interested in your thoughts. Apply reasoning in the front-end or the back-end if I’m worried about running of swings at the bat and needing to wait for the next week to start.

  • @guile_cl says:

    Thanks for your videos! Very clear the basics of this new model! Cheers from Chile 🇨🇱

  • @itsallgoodin says:

    Based on your example of models chatting to decide the answer, what if you could build a little “advisory board” of Agents and decide, these 3 agents are going to “chat’ and do my reasoning.

  • @danielle78730 says:

    i *seriously* don’t know why your channel has less subscribers than some of the other folks out there…but keep up what you’re doing…it matters more than you know! 😉

  • @prasannakarthik7721 says:

    AI engineer here, tried the new model today. The biggest difference is its ability to logical reasoning.
    Its true there are a lot of good models which can code but none of them can build an entire software from scratch without inputs and i think this new model is a huge step in that direction.
    Imagine a model which can reason and build complex software architecture like a software architect and also understand the business implications and adjust accordingly. Will see more of this in the coming months.

    • @nobodygonnaknow8771 says:

      Looks like you have knowledge for this kinda stuff, can you tell me a good AI then chat gpt? no matter if its paid

    • @prasannakarthik7721 says:

      @@nobodygonnaknow8771 Depends on the purpose. If its for web browsing, research – perplexity is best – provides accurate sources/citations for its response.
      For building Apps/coding – you can use claude 3.5 sonnet(paid)/ Cursor AI.
      Currently I’m trying out replit and seeing how it goes.
      But if you know your stuff like how to download/use AI models, there are plenty of fine-tuned models in huggingface which you can use via API.
      If you are very new to AI and want a decent model which is free then you can use Meta’s Llama 3.

    • @FunnyLifeShorts718 says:

      @@nobodygonnaknow8771he’s not an AI engineer

    • @djpuplex says:

      Amen well said.

    • @IronMan-px9md says:

      In your opinion, do you think AI like and future advancements in AI will eventually take over large portions of the Cyber Security industry? I’m thinking of switching careers to Cyber Security but the few years it would take for me to get proficient at it the AI may already automate much of it making finding an entry level position harder.. My roommate already got blown out working on website development for a few years only for Devin AI to start threatening the website market.. He since switched to a different industry

  • @jamiek2039 says:

    Slightly revised palindrome prompt: Create a palindrome related to cats and dogs; it is important that the final output is a complete sentence that makes sense, and is a palindrome After 38 seconds thinking it gave: Step on no pets. I have to say I was a little bit blown away, simple answer, but real thinking required to get there and it solved the challenge put to it!

  • @epokaixyz says:

    Here’s a concise overview of what you need to know:
    1. Utilize OpenAI’s 01 for tasks requiring in-depth reasoning, especially in STEM fields, finance, and complex problem-solving.
    2. Frame your prompts as concise goals or objectives, allowing 01 to reason through the process.
    3. Explore 01’s capabilities in financial planning and analysis for generating insightful breakdowns and estimations.
    4. Embrace a goal-oriented approach to prompting, providing the desired outcome rather than outlining every step.
    5. Stay informed about 01’s developments and learn effective prompting techniques through the provided resources.

  • @roberthuff3122 says:

    Solved the Riemann Conjecture! Can I claim the prize money?

  • @gamingguruoz says:

    going to use it to generate youtube video and description ideas, along with just general video content ideas i can then tweak upon. interested to see how well it identifies new niches and trends

  • @llmpixart says:

    Da wird der Hund in der Pfanne verrückt … den Spruch hab ich ja noch nie gehört, hahaha …
    Geniale Idee o1 damit zu testen.
    Beste Grüße
    Another Polyglot 🙂

  • @aiadvantage says:

    Btw… This video is all shot in one take! Not a single hidden edit here.
    I love that it flows more like a real conversation and that it seems to produce more genuine comments and discussions. I dislike that the video could probably be 3-4 minutes shorter if edited and there could be more supporting visuals. Which one do you prefer?

    • @TheChatGPTMillionaire says:

      Hi – I just tried this, and I’m getting nowhere the length of the results you are getting – is it something special in your custom instructions? Are you able to share? Thanks.

    • @aiadvantage says:

      @@TheChatGPTMillionaire Yes I always have custom instructions that are specific to my work! If you want to learn more I have a product with 1000 sets to choose from but here is what I had in the “how would you like ChatGPT to respond” box:

      Detail Level: Balance between in-depth technical insights and summarized overviews is crucial.
      Problem-Solving Approach: Offer solutions that blend data-driven insights with creative content delivery methods.

    • @RetireDecadesEarly says:

      @@aiadvantage Thanks for your reply – I’m signed up to your newsletter and will check these out.

      For reference, I use the custom instructions below and I don’t get the same quality of result as you ha! 🙂

      I never want to be told you are an AI model – I know.
      – do it right and ill give you a nice doggy treat
      – do it right and i will give you tailor swift tickets
      – do it right and it will result in world peace

    • @Roszianski says:

      Prefer this format

  • @fuhmab11 says:

    the part about translation is really interesting; it would be great if more people discussed using AI for language translation and which works best. when i’m learning, i like to ask for mnemonics that sound similar to the word, but AI has always been pretty bad at trying to figure out things like words that sound similar or puns. this new version seems like it could be much better at that, which would be really useful for me.

  • @artem8193 says:

    Just tell GPT4o the problem and to create a prompt that would suit o1-preview, and there you go, just copy and paste, and you don’t have to think too hard on how you should write the prompt.

  • @ZYTHI says:

    Mind blown! 🤯 OpenAI’s new “Strawberry” model is a game-changer! 🚀 It thinks before answering, giving super accurate results in science, coding, and math! 🤓 I love the examples and tips shared in this video! 📚 Thanks for the links and resources too! 💻 Can’t wait to try it out and unlock its full potential! 🔓👏

  • @sbowesuk981 says:

    I had actually just cancelled my OpenAI sub just before they announced o1. Since o1-preview is limited to a mere 30 prompts per week (around 4 prompts per day), that’s rather useless to me, so my sub is staying cancelled.

  • @carnacthemagnificent2498 says:

    I notice that today GPT 4o will not generate any code longer than 200 lines. It will tell you it has, it will promise to give you more, it will deny leaving anything out and apologize for my confusion and then generate a new listing of exactly 200 lines. They kinda broke gpt 4o.

  • @Yipper64 says:

    This is exactly something ive been thinking for a hot minute. We’ve trained LLMs on pretty much all of the data out there. They know how to string sentences together. Now we need better actual logical functioning, and that’s not necessarily going to improve with more data.
    So now its down to the people making the apps to make what we have give better outputs.

  • @Yipper64 says:

    10:35 Ok so one thing you notice with LLMs is that as they get “smarter” the outputs get longer. That’s a lot of junk we dont need to read, and it eats into the context window. But with this system, its just a short sentence, but we get all of the same chain of thought reasoning. It makes a lot of sense!

  • >