- CaffeinatedGPT
- Posts
- š° Bill Gatesā Thoughts on GPT-5
š° Bill Gatesā Thoughts on GPT-5
DALLĀ·E is Now Available in ChatGPT

3 Min Read ⢠Intensity: āļøāļø
Good morning, Caffeinators āļø
Today, we present you with a conflict between OpenAI and its stakeholders (mainly Microsoftās big boss).
Starting Sips:
š¬ Bill Gatesā Thoughts on GPT-5
šØ DALLĀ·E is Now Available in ChatGPT
šŖ OpenAI Aborts Secret Plan - Arrakis

Yes. Bill Gates is making predictions about GPT-5. And it is not a pretty one.
He believes that we might've hit a ceiling with current GPT tech. And, GPT-5 wonāt be that much of a breakthrough.
Specifics:
Bill Gates suspects GPT-5 wonāt be as good because OpenAI will spend more time understanding the current tech
OpenAI CEO Sam Altman and others disagree: GPT-5 will be significantly better than GPT-4
Both parties agree that the next big goal is to cut down on AI chip cost
My Thoughts:
Understanding AI's logic is not just for tech's sake. It has ethical implications. Humans need to ensure AI's decisions are justifiable.
Even though OpenAIās core tenet is to develop AGI, Microsoft holds 49% of the company. In other words, Bill Gates has a huge influence.

GPT has upped its game! You can now create ANY images using DALLĀ·E.
Simply describe a thought, and DALLĀ·E 3 shall conjure.
Details:
Create a photo-realistic picture of nature, a pet store website mockup, a business logo about fruits, and more!
DALLĀ·E is better now. It is the best tool for AI text generation in images (see image above)
DALLĀ·E uses GPT behind the scenes, so it understands you better than any other AI tools
*The tool is available for Plus and Enterprise users
What Matters:
It is exciting to see how powerful DALLĀ·E is today. But, we also should be worried that this powerful tool is available to everyone.
Thatās why OpenAI is taking active measures to prevent bad things from happening, including:
The model avoids violent, adult, or hateful content through safety checks (Iāve also tried generating real people and failed)
Itās trained to prevent mimicking living artistsā styles
OpenAI can now classify if an image is AI-generated with up to 99% accuracy
If you want to learn more about DALLĀ·E, stay tuned for the next issue!

OpenAI started a secret project called Arrakis. The main goal was to make AI computing cheaper.
However, OpenAI had to call it quits early because of quality concerns.
Main Points:
Arrakis is supposed to be better because of the sparse principle, where only parts of the neural network are used (unlike traditional dense models)
OpenAI faced challenges with performance. So they halted the project even though it showed promising results
OpenAI is now focusing on building a faster GPT-4 version
My Thoughts:
OpenAI's venture with Arrakis showcases the unpredictable nature of AI development. Even with promising starts, outcomes can be unexpected.
Most importantly, it is great to see that OpenAI still operates like a startup. Even though Microsoft and other stakeholders had high hopes for Arrakis, because of how much money they could save, OpenAI was able to make the final decision and shut the project down.
š„¤ To-Go Cup
š¤ DALLĀ·E, an AI image generation tool, is now available for ChatGPT Plus and Enterprise users
šØāš¼ Bill Gates thinks GPT-5 wonāt be as big of a breakthrough as GPT-4
šµ OpenAI aborts Arrakis, instead focusing on a turbo GPT-4
If you enjoyed this article, I think you will like this one too! It has 10 examples of using GPTās vision (my favorite is the dog one).
Until next time,
Your friendly barista and āļø GPT!
Reply