As a preface, I am not an expert in AI in any capacity, I just think it’s really cool! Take everything here with a grain of silicon.
Artificial Intelligence. ChatGPT-4. Robot apocalypse. Social media has exploded over the last few weeks with ongoing developments in the development of artificial intelligence – and a lot of people have questions.
So what’s an AI? Is it alive? Is it a robot? Is it going to destroy humanity?
No! Well, not yet! AI stands for Artificial Intelligence. For centuries humanity has been fascinated by the concept of making robots, automata and minds that are not human but think like humans. These minds have always been a product of science fiction or fantasy. Day by day, we seem to be getting slightly closer to actually achieving those dreams. It’s not a robot, like R2-D2 or Daleks or Cylons; it’s not capable of actual independent thought. As far as destroying humanity, let’s say… probably not!
So, like, what actually is it? How does it work?
Current AI are mostly predictive language models. There’s not actually a mind in there yet – instead, it’s much closer to a souped-up version of the autocorrect feature on your computer or phone. AI like ChatGPT consume huge amounts of data, looking for patterns and repetitions in the way language works. When you ask it a question, it produces answers that it thinks probably would be given to that question. While the answers it gives can seem freaky or really close to those a human might give, it’s important to understand that it’s generating those answers based on the information it’s been given, not truly thinking in some odd alien way.
What can it do?
At the moment, AI are mostly split between limited industrial ones that automate basic processes, and experimental ones (like ChatGPT, Bing, Midjourney, and lots of others). Industrial AI are used in a lot of behind-the-scenes processes at major companies to solve problems that have easily predictable solutions, but tend to take a lot of time to solve when a human does it. This includes things like solving logistics obstacles (scheduling, efficient movement of goods, shipping routes, and much more) and mathematics problems that take a huge amount of effort to calculate. Another area these models are deployed is in data analysis. Companies are gathering a lot of information from customers, and realistically are unable to process a large portion of it in any efficient way – which is where AI come in.
What problems does it face?
On the practical side, hallucinations are one major trouble of this particular style of AI. When asked a question it has no data for, these language models are known to occasionally answer very confidently with completely wrong answers. The machine seems to decide a random answer is best, and doesn’t seem able to recognize that it’s giving a wrong answer at all. This is a particularly prevalent problem that will need to be solved before AI can properly develop any spark of sentience – after all, a major part of communication is the ability to simply say “I don’t know” instead of lying. This is a great reason to, say, ask a reference librarian instead of an AI if you have a question! (Don’t outsource me yet!)
Ethically speaking, the entire field of AI opens endless new questions. The easy questions are ones we’ve been asking for centuries – just look at Mary Shelley’s Frankenstein! Is it right to create something that can sort of, kind of, think by itself? How do we teach it to act ethically? After all, chatbots such as Tay and Blenderbot 3 already demonstrated that, if given bad inputs by users, the outputs will become awful. The hard questions move on to more practical concerns – if, as OpenAI’s research suggests, nearly 80% of the workforce will see their jobs impacted by AI in some fashion, what does that mean for those of us who need a job to survive? What happens when automation replaces… everyone? Don’t panic quite yet! These questions will need to be solved on a very large scale – you’re probably not going to be replaced by a computer any time soon.
Ultimately, for every new answer we get through these experiments in artificial intelligence, we create a thousand new questions. Unsatisfying as it is to say “I dunno” to most of these questions, whatever comes next is going to be incredibly important. Maybe just keep a glass of water near your computer for the moment.
-Christian Boor, Digital Services Librarian
