- RuntheAI
- Posts
- GPT-4: What we know
GPT-4: What we know
RUNTHE AI
December 25, 2022
Good Morning AI Runners
Here's what we've got for you today:
GPT-4: What we know
ELI5: Chinese room argument
GPT-4: What we know
There has been a lot of discussion about GPT-4 on Twitter, including a thread about its potential capabilities and the use of supercomputers for its development (linked below).
But how different is GPT-4 from GPT-3? what can it do differently? and when should we expect it to be released?
GPT-4 is expected to be a major upgrade from GPT-3 and is rumored to have 1T parameters, which is six times larger than GPT-3 (175 billion parameters).
GPT-4 will be significantly more advanced than GPT-3 and may be able to convince many people that it is truly intelligent. However, it is not expected to be a true artificial general intelligence (AGI) and may still have limitations in its understanding and processing of complex information.
Side note: AGI is basically "Artificial General Intelligence." It is when computers are able to think and learn in a way that is similar to a human. This means that an AGI would be able to understand and solve a wide range of problems, not just ones that it was specifically programmed to solve. It would be able to learn and adapt to new situations, just like a person can. AGI is still a very new and developing field, and it's not yet possible to build a computer that has AGI.
Some predictions based on rumors floating around:
Biggest model size for GPT-4 will be 1T parameters. Up 6x.
OpenAI launched ChatGPT to identify instances of potentially misleading information and to test safeguards to optimize them for GPT-4.
Clearing Turing test and any standard tests.
If you are interested in learning more, here's an in depth thread on supercomputers and GPT-4:
There's been a lot of low quality GPT-4 speculation recently. So, here's a relatively informed GPT-4 speculation thread from an outsider who still doesn't know that much. 🧵
— Matthew Barnett (@MatthewJBar)
10:26 PM • Dec 20, 2022
Chinese Room
The Chinese Room Argument is a thought experiment proposed by philosopher John Searle in 1980 to illustrate the differences between human and machine intelligence. It involves a scenario in which a person, who does not understand Chinese, is given a set of rules to follow in order to respond to Chinese writing. The person follows the rules and writes back in Chinese, but does not truly understand the meaning of the language. According to Searle, this is similar to how a computer operates when it follows a program. While a computer can perform tasks that seem intelligent, such as understanding language or solving problems, it is merely following a set of rules created by someone else and does not have true understanding or consciousness.
The Chinese Room Argument has been used to argue that computers can never truly be intelligent or have consciousness in the same way as humans. However, some argue that the argument does not take into account the capabilities of modern artificial intelligence systems, such as their ability to learn and adapt based on experience. Overall, the Chinese Room Argument is an important consideration in the field of artificial intelligence and our understanding of AI.
Run The AI: PPPs
Pick up (learn):
The TOP500 project ranks and details the 500 most powerful non-distributed computer systems in the world. Currently at number 1 is the Frontier machine at Oak Ridge National Laboratory. Checkout the rest of the list below 👇
Pilot (play):
Generate a presentation with AI powered tome:
Person:
Alexandr Wang, cofounder and CEO of Scale Al.
Scale delivers high quality training data for AI applications such as self-driving cars, mapping, AR/VR, robotics, and more.
Pic of the day:
That's it from RuntheAI for today.
THANK YOU FOR READING AND SEE YOU TOMORROW, SUBSCRIBE TO STAY UPDATED!