“The Future of AI” with Michal Kosinski | All Else Equal
In an insightful discussion on the future of AI, Professors Jonathan Berk and Jules van Binsbergen, along with Stanford Professor Michal Kosinski, delve into the potential impact of AI on businesses, the progress of AI models in understanding and predicting human behavior, and the need for regulation in this rapidly advancing field.
AI’s Problem-Solving Abilities
AI models, such as GPT-4, have shown remarkable progress in solving complex tasks that require a theory of mind.
These tasks, which were previously challenging for AI, are now being solved at a level beyond human capability, showcasing the rapid evolution of AI technology.
AI Outperforming Humans
AI models have been tested on various psychological and psychometric tasks, and have shown a clear progression in their abilities.
In some tasks, such as the SATs and bar exams, AI has even outperformed humans, indicating the potential of AI to perform intelligent tasks.
Emergence of Unexpected Properties in AI
As AI models are trained to be more human-like, they can develop unexpected properties, such as biases.
This raises questions about what other human traits AI models might develop in the future, and how these can be controlled or understood.
I was stunned how suddenly just from one version to another, this large language model went from just a complete idiot when it comes to this particular task, to a genius, where they essentially started solving all of those tasks correctly. – Professor Michal Kosinski
AI: A Comparison with the Human Brain
AI models should not be viewed merely as advanced tools.
Instead, they should be compared to the human brain, suggesting that anything a human brain can do, a computer program could potentially do as well.
Need for Checks and Balances in AI
Just as systems have been designed to limit human power and influence, similar systems should be implemented for AI.
With the potential for AI to be manipulated and misused, the importance of safeguards and regulations cannot be overstated.
AI’s Limitations
Despite its impressive capabilities, AI has its limitations.
For instance, while it excels in language prediction, it lacks the ability to manipulate the physical world or perform calculations like a calculator.
AI as Storytellers
AI models are more akin to storytellers or improv artists than fact-checkers or calculators.
They excel at continuing a given sentence or narrative, not necessarily at providing factual information.
Differing Views on AI’s Potential Dangers
While some believe in human rationality and the ability to adapt to AI advancements, others express concern about AI’s potential to manipulate group dynamics and influence large groups of people through selective information.
The Uncertain Future of AI
Predicting the impact of AI is challenging due to its disruptive nature.
However, it is crucial to continue the conversation and explore possible regulations and safeguards to mitigate potential risks.
