Previously I wrote about ChatGPT and what they are doing to change the face of artificial intelligence. I will now take that a bit further and write about artificial general intelligence or AGI and what it is capable of. AGI can already exceed what humans are able to do intellectually speaking according to an article on https://spectrum.ieee.org/superintelligence-christoph-koch-gpt4. While that is pretty impressive it is also a bit frightening.
OpenAI is a research organization that has produced the vastest learning language models (LLMs) for AI and they feel that this technology is moving rapidly forward. A certain amount of misinformation is also being posted and that needs to be quashed. I will attempt to keep this as close to the facts as possible.
According to Christof Koch of Seattle’s Mindscope Program at the Allen Institute AI is inappropriately defined and as such most people lack an understanding of it. To which he adds, “Intelligence, most broadly defined, is sort of the ability to behave in complex environments that have multitiudes of different events occurring at a multitude of different time scales, and successfully learning and thriving in such environments.”
Regarding AGI Koch says “…it would be like an intelligent human, but vastly quicker. – you can ask it any question, and you immediately get an answer, and the answer is deep. It’s totally researched. – you can ask it to explain why…” This he says is what ChatGPT is capable of. That is incredibly impressive if you ask me. Furthermore, Koch says “…you can ask it to write code, – it can go through the program, line by line, or module by module, and explain what it does. It is a train-of-thought type of reasoning that’s really quite remarkable.”
I am not a coder, but would love to be able to understand that intricate process, and if ChatGPT can enlighten me in a regular person language then bravo to them!
According to the website https://www.medicalnewstoday.com/articles/321060#_noHeaderPrefixedContent “The brain is the most complex organ in the human body. Many believe that a person only ever uses 10 percent of their brain.” To which they add, “A person’s brain determines how they experience the world around them. – …contains around 100 billion neurons – cells that carry information.”
Medical News Today says that the 10% rule is a total myth. If we want to be capable of more with regard to our brains we must feed them. Brain function is tied to numerous things beginning with eating a balanced diet and getting regular exercise. The excessive use of alcohol can also lessen our mental processes greatly.
“Most people think about AGI in terms of human intelligence, but with infinite memory and with totally rational abilities to think – unlike us. We all have biases. We’re swayed by all sorts of things that we like or dislike, given our upbringing and culture…AGI would be less amenable to that.”
AGI is likened to superintelligence. Superintelligence is genius-level intelligence that is very very fast and extremely accurate. Most regular people do not process information at this level. AGI is capable of reading all of the books on all of the topics that are available to the world today and is able to siphon through all of that information to respond to our questions at large. To wit, “…although it wasn’t trained to write love letters, it can write love letters. It can do limericks. It can generate jokes.” It can also create trivia questions, write code, and is fluent in all languages. That is amazing!
Koch says, “ChatGPT reminds me of a widely read, smart, undergraduate student who has an answer for everything, but who’s overly confident of his answers and, quite often, his answers are wrong. – You can’t really trust it.”
According to Koch we need to understand the distinction between consciousness and intelligence to move forward. They are not the same thing. “Intelligence ultimately is about behaviors, about acting in the world. If you’re going to do certain behaviors and you’re not going to do some other behaviors. Consciousness is more a state of being. You’re happy, you’re sad, you see something, you smell something, you dread something, you dream something, you fear something, you imagine something. These are all conscious states.”
With great promise also comes, perhaps anyway, great threats if this technology is misused. Koch said, “It’s not consciousness that we need to be concerned about. It’s their motivation and high intelligence that we need to be concerned with.”
Koch entertains one of two theories of consciousness, the Integrated Information Theory. The premise of this theory is that we cannot forge consciousness. “ …you will be able to build a computer that simulates a human brain and the way people think, but it doesn’t mean it’s conscious. – Just because you can simulate the behavior associated with consciousness, including speech, including speaking about it, doesn’t mean that you actually have the power to initiate consciousness. – these computers, while they might be as intelligent or even more intelligent than humans, they will never be conscious. They will never feel.”
Koch says that feeling is not an important attribute to the practical applications of AGI. Instead, we need it (AGI) to be able to reason and predict actions based on that reasoning. People are the most prevalent species on earth and also the most aggressive. Technology is closing the gap between machines and men. Machines may soon possess the ability to be both intelligent and aggressive. As a society, this is what we should be concerned about most in this regard.
Koch uses Tesla automobiles and what they are currently capable of as an example and then combines that ability with what ChatGPT and drones are able to do. Then he adds, combine those qualities with drones being able to access the cloud and obtain more intelligence and reasoning. “…then they begin to understand our motivation – including maybe not just generic human motivations, but the motivation of a specific individual in a specific situation, and what that implies. – When people say in the long term this is dangerous, that doesn’t mean, well, maybe in 200 years. This could mean maybe in three years, this could be dangerous.”
With the inception of incredible information at a keystroke there is also as much, if not more, misinformation that would be available. We have already seen some of this in our regular and social media. What can we do?
Koch says there is a motion being considered at this time from the Future of Life Institute that is asking the tech sector to take a time-out for approximately a year and a half before progressing with regard to the newest learning language models. He is cautious to state that while this is being bandied about it is not a call to halt all development of such technologies, but a point at which we might consider how best to safeguard against the potential abuse of such technologies. Specifically, the direction we want to take in this regard for the near and far term. “Do you think the upsides are going to outweigh whatever risks we will face in the shorter term? In other words, will it ultimately pay off?”
Koch says there are two lines of logic with regard to this time-out and they begin and end with those engaged in this AGI process. AGI has put highly talented programmers at the center of this dynamic. They are well-paid for their services and totally committed to enlarging their footprint, but can they be trusted to not abuse it?
For some, this is the stuff of science fiction. For others, those in trenches that work with it, it is their lifeblood. It is the reason they go to work each day.
To date, we are not able to get AGI to write a novel on the level of War and Peace, but most feel that doing so is merely a matter of time. As with all things, there are good and valiant actors and there are actors that will seek to manipulate this technology in their favor. Either way, this is our current world. Tread carefully.