Marci Harris, French teacher, thinks that Al won’t be taking over jobs any time soon. She thinks that they will be helping professionals improve their work rather than replacing them. Personally, she uses Claude to help find her starting point in writing letters of recommendation, and she’s able to modify and personalize the work afterward.
“You know how when people ask someone something they don’t know, they’ll say like, ‘oh, go google it, and I think that’s what it’ll be like with AI,” Harris said. “There’s a saying that Al won’t replace you, but the person using it will. I think we all can benefit from using it.”
Recently, within the world of tech-nology, there’s been a huge Artificial Intelligence (AI) boom that has come with rapid progress in just the past few years. Let’s take, for exam-ple, ChatGPT: the generative Al that we all know for doing homework, answering questions and even generating images. Launched in November of 2022, the Al was text-only and often made up facts when it didn’t know the answer, later being defined as “hallucinations.”
Two years later, it’s now more fluent at responding — but still has the “hallucinations” quite often – and can now create photos and art from a prompt and analyze photos. The company that made ChatGPT, Ope-nAI, has been developing a text-to-video Al model called Sora, which is currently not open to the public.
Looking at face value, it seems like everything is only improving exponentially – but that’s not exactly the case. Most professionals believe that growth follows an S or sigmoid-al curve: it has a shallow developing phase, a steep growth phase and a maturity/limit phase that flattens out, and we happen to be on the very steep part of that curve right now.
“The biggest foundational enabling piece that allows technology to advance so fast right now is really a wealth of data that was previously unavailable,” Jose Nazario said, who works in cybersecurity. “Before, there were teams of people who are basically teaching computers what math is and how it fits together, and they’re doing this manually. Now, like OpenAI, Google, Microsoft and others really have the world’s data available to them, and very large computers to process it.”
Nazario works in cyber threat intelligence, meaning he investigates cyber attacks and gathers data on who’s behind them to try to understand what happened and what matters. He uses this information to warn and inform his customers. Recently, the two big areas he’s worked on are the Russia-Ukraine conflict and the Israel-Hamas conflict.
He uses Al to accelerate parts of his work. For example, translating, automating and structuring data, although he still finds that it does a pretty terrible job at replacing analysts and their unique insights due to the fact that they are rare and require creative thinking, the two places where Al really falls.
“I believe we’re already seeing outsourcing by some firms to automated software development platforms hosted by companies like GitHub and others. The quality of the code, the security of the code, and things like that are open questions, so that gives some of my peers and colleagues a lot of concern about this situation,” Nazario said. “It could just be a reaction to people feeling like their jobs are under threat, but I think we’re already seeing some of this happen. I would argue the story of John Henry is instructional here: Remember, in the story, John Henry swings a hammer to lay the railroad, and he’s challenged by a steam drill. He fights back, but dies in the end.”
Currently, well-known AI – ChatGPT, Gemini and Copilot just to name a few – doesn’t really have “intelligence.” They aren’t really original thinkers, but more like noise machines that generate answers based on what they’ve been taught. They aren’t creative and they’re unable to create their own original work.
We haven’t been able to make AI with human-like intelligence and the ability to self-teach, also known as artificial general intelligence (AGI), and I don’t think we will be able to in the near future. Right now, most Als are made to do specific tasks such as playing chess, chatting or generating art and we would need to combine them all. And with what professionals have learned they don’t generalize well.
“Think Ultron, for example, a machine that can really sort of master domain and its own objectives to achieve them,” Nazario said. “That is, in my estimation, preposterous right now to imagine having any real chance of happening, but that also doesn’t necessarily stop people from throwing money at it.’
Right now, many of the big companies who have a large share of the market, such as Tesla and OpenAI, are working towards AGI as sort of their goal, or promise. So what happens when these big companies with a lot of funding and hype fail? Well, there’s been a specific term coined called Al winter, which in a classic case is where almost everybody put away their research programs, fund-ing, hopes and dreams for Al for a long time.
We probably won’t see something on this level and quite as abrupt as this, but there certainly will be a reckoning. As it is right now, the promise and hype which back the funding and investments seems to outstrip the real promise here. Just this past summer, there has been a report made by Goldman Sach that looked at the anticipated payoffs of Al. It noted that the numbers that people were promoting and promising are simply unobtainable and don’t justify the levels of investment that we’re seeing.
“It’s already happened at smaller scales, where we have seen venture investments get a lot more savvy and scrutinize deals more, the land grab, if you will, is shifting into a new phase,” Nazario said.
So maybe not a full “winter,” but a cooling off of sorts seems likely.
Even though Al development is fast right now, the future of robots taking over the world doesn’t seem too close in grasp. However, word is that OpenAl is currently working on a secret project called Project Straw-berry, which is supposed to push the boundaries of Al reasoning and be a step towards the AGI enthusiasts want to see, but only time can tell if it turns out well or not.
“I don’t have a crystal ball, but this sort of hype cycle has played out before to these kinds of effects,” Nazario said. “Maybe it’ll be more like the Internet revolution wise at the turn of the century, where there was a lot of investment, a lot of silly money thrown in, a lot of things that didn’t pan out, but ultimately, there were some lasting, durable changes that could get built on.”
