Numerous specialists who have commented on the situation agree that Lemoine was deceived. LaMDA does not feel like a person just though it speaks like one. But, the revelation raises fears about the future. As AI becomes sentient, we must understand what sentience is and how to test for it.
It is commonly known that AI can tackle issues that would ordinarily need human intelligence. Yet, “AI” is a wide word that may refer to a variety of systems, according to Sam Bowman, an AI researcher and associate professor at New York University. Many variations are as simple as a chess software on a computer. Others include sophisticated artificial general intelligence (AGI) – systems capable of doing every task that a human brain can. Some advanced versions include artificial neural networks, which are algorithms that loosely replicate the human brain.
“As more and more artificial intelligence is entering into the world, more and more emotional intelligence must enter into leadership”
LaMDA, for instance, is a neural network-based large language model (LLM). LLMs compile text in the same manner as humans do. They don’t, however, just play Mad Libs. Linguistic models may also learn to translate languages, have discussions, and answer SAT problems.
These simulations can fool humans into thinking they are intelligent long before they reach that stage. After all, engineers created the model to mimic human speech. If a person claims to be sentient, so will the model. “We can’t rely on self-reports for something right now.”
According to Long at Oxford, large language models are unlikely to constitute the first conscious AI, notwithstanding their ability to trick us. AI systems that learn for lengthy periods of time, perform different jobs, and defend their own bodies, whether actual robot encasements or virtual projections in a video game, are more likely prospects.
Alan Turing proposed the “imitation game,” sometimes known as the Turing test, in 1950 to determine if machines can think. An interviewer conducts a conversation with two individuals, one human and one computer. If the computer continually convinces the interviewer into thinking it is human, it will pass.
Researchers now believe that the Turing test is an inadequate measure of intellect. It evaluates how successfully machines fool people under simulated settings. Computer scientists have progressed to more complex assessments, such as the General Linguistic Understanding Evaluation (GLUE), which Bowman contributed to the development of.
Even if it isn’t conspiring to take over the world before the hero arrives to rescue the day, AI has a cool factor. It appears to be the type of tool we want to delegate the heavy lifting to so we can do something entertaining. Yet it may be some time before AI, sentient or not, is ready for such a massive leap