How worried should we be about the AI takeover?
If you’ve been reading any articles lately, you’ll notice that a great many of them are talking about the rise of Chat-GPT and AI in general. If you’re like me, you may start to wonder how much this should worry you. Will my job be in jeopardy? Will AI come for all our jobs? You might even have a little bit of anxiety about these questions. Let’s go through it together and dissect what AI is and isn’t, what AI currently does, and what AI is projected to do. I think you’ll find that we shouldn’t be as worried as everything is trying to have you believe we should.
What is AI today?
Today, AI mainly comprises various types of Machine Learning models. These are algorithms that can “learn” based on training data sets. Most models are trained for a very specific purpose. For example, some are trained to turn text into speech, some to turn images into text, and some to translate different languages. Most models you’ll encounter in the wild are trained only to do a specific thing. However, there are a select few that are slightly different. Chat-GPT is one of those types of models. Chat-GPT is a large language model, which means that its training data was essentially all the data on the internet (that OpenAI could collect, at least). You can read about how they trained Chat-GPT here: https://openai.com/blog/chatgpt/. I’ll follow up with an article about Chat-GPT and what it can and cannot do.
The important thing to note about Chat-GPT is that it is a unique case. Most models cannot do everything that Chat-GPT can. Since its release, companies like Google have been scrambling to match it but have fallen a little short. This should tell you how difficult it is to make something like Chat-GPT. Even today, Chat-GPT has its flaws, as do most models. To summarize, today AI is made to solve specific problems and automate specific tasks. Models that can do everything all at once are far from reality.
What AI isn’t
You’ll hear a lot of nonsense about AI taking everyone's jobs and putting people on the streets. At the time of writing this, that is just not true. One of the most common types of AI people talk about, in terms of putting people out of work, is generative art AI. There are a lot of models out there that can take a prompt and turn it into art. DALL.E 2 is one good one that you can play with: https://openart.ai/. Many article attempt to convince that this will put artists out of a job. If you scroll through some of the creations, you’ll see some really stunning images that it created and the prompt that was used. At first glance, it can seem like this could replace human artists.
However, there are some real issues with generative art models. For one thing, they cannot create “unique” art pieces. These models are trained on billions of pieces of captioned art. So when you give it a prompt, it’s using the art it has seen to create a sewn-together knock-off. It has no thoughts, no feelings; it just spits out derivative art pieces that sometimes come out a little wonky. Another issue is that it does not understand the context of your prompt. So if you ask DALLE.2 to create an image of a recent event, it won’t really have that context and won’t give you a precise piece of art.
This is just one popular example. There is still a great market for artists, even though models like DALLE.2 are becoming quite popular. Although this is only one example, every industry can take a deep breath because AI is very far from replacing human workers.
The Future of AI
Now that we have a bit of context about the current limitations and uses of AI, where is it going? Some would have you believe that AI will eventually replace you in the workplace, putting you out on the street. I think that’s an overdramatic take. I think in the immediate future, we’re going to see AI become just another tool in our belt. This is already becoming a reality with things like GitHub Copilot.
Image taken from original article: https://github.blog/2022-06-21-github-copilot-is-generally-available-to-all-developers/
I’ve been using Copilot for about a month now, and I’ll give you more specific thoughts in another article. However, what I will say now is that it’s a pretty good tool. I think most AI tools will begin to augment tasks rather than replace the human operator. Another example of this type of AI integration is in medicine. There are plenty of AI models used in medical imaging to help detect and predict the growths of cancer cells/tumors. To be clear, these are incredibly important tools that we should be investing in.
What about the distant future? I’m not going to make any specific predictions about the distant future of AI because it’s almost impossible to predict where we’ll end up in this regard accurately. Some believe we will merge with AI becoming cyborgs. If you look at companies like neuralink, the Elon Musk-funded company attempting to implant chips in people’s brains, you may think there is a lot of investment in that area (I’m not going to delve into the MANY issues that neuralink has, we’ll be here all day). There are also people who are implanting themselves with chips and things to enhance their own senses, as well as people who need surgeries to help them live, like prosthetic body parts.
When it comes to AI, I think there are a lot of questions that need answering. I think there need to be more talks about rules and regulations about what we can and cannot do (like killer robots would be a big nono for me). I am generally optimistic about the future of AI in that it will be a tool to help us do more research and become more productive while giving us more freedom as well. If you read articles that try to sell you impending doom and gloom, know that’s probably not as bad as they’re making it seem.
