Surely if you’re reading this, in the last few days, weeks, or even months, you’ve heard about artificial intelligence. This is a very interesting topic that has been gaining more and more relevance lately. In this article, we’ll take a journey through artificial intelligence — we’ll talk about its history, its present, and what the future may hold.
I should point out that the cover image of this article is not a photo taken by a person or a drawing by someone — nope! It’s an image generated by DALL·E 2’s artificial intelligence, which we’ll talk about later.
What is Artificial Intelligence?
Before we begin, we need to understand what artificial intelligence is. The Royal Spanish Academy defines “artificial intelligence” as: “a scientific discipline concerned with creating computer programs that perform operations comparable to those carried out by the human mind, such as learning or logical reasoning.”
Honestly, the RAE gives a pretty accurate definition of what AI is. But let’s try to understand how it works.
Learning in Artificial Intelligence
Let’s think about our own intelligence, human intelligence. We constantly receive inputs, which we’ll call E. These could be problems we need to solve, messages from friends, or a lecture from our favorite teacher… In response to these inputs, we, as intelligent beings, need to generate responses (called R), responses that aim to get as close as possible to the best possible solution — the supreme answer that solves the problem (if it has one).
After receiving E and performing R, we get an output (S). If S matches the intended goal of E, then we remember R as the correct response to E. Otherwise, we try a different R until S matches the goal of E.
Let’s use an example. E is “Watch Formula 1”. We, as F1 fans, get this urge. Then we try to respond with R. R is “Open the window”. After opening the window, we evaluate whether S, which is “Window opened”, corresponds to E. The answer is clearly no — opening the window won’t let us watch Formula 1 (unless a neighbor is watching it and we can see their TV). Since S doesn’t match E, we try another R. “Turn the front door key” is our new R. This gives us “Door opened” as S. Again, we ask: does S match E? No. So we keep trying until S finally matches E. This trial-and-error process is one way AI learns.
Another way is by being taught. If someone teaches me that to watch Formula 1 I need to turn on the TV and tune to the right channel, I save that in my “hard drive.” So when I get E, “Watch Formula 1,” I respond with R, “Turn on the TV and tune to the right channel,” and get S, which now matches E.
The Past of Artificial Intelligence
If there’s someone who played a major role in what we now know as AI, it’s Alan Turing, who in 1950 wrote the article “Computing Machinery and Intelligence,” in which he asked whether machines can think. In that same article, he proposed the famous “Turing Test”, which is a test to determine whether a machine can exhibit behavior indistinguishable from a human. The test involved a machine and a human answering questions from a judge via a monitor. If the judge couldn’t tell who was the machine, the test was passed; otherwise, it wasn’t.
Alan Turing (https://www.elcorreo.com/).
However, many researchers and scientists claim the real starting point of AI was in 1955, when John McCarthy, Marvin Minsky, and Claude Shannon met at the Dartmouth Conference, where the term “Artificial Intelligence” was first introduced.
Later, in 1977, a computer defeated world chess champion Garry Kasparov for the first time, thanks to IBM’s supercomputer Deep Blue.
In the 2010s, AI reached new heights with Apple’s Siri, Google’s Google Now, and Microsoft’s Cortana — voice-powered virtual assistants.
Thanks to all these milestones, we now know AI as we do today, with the following applications.
Today’s Artificial Intelligences
OpenAI
This company researches AI development and aims to create new AIs. In fact, it has developed several well-known AIs lately, such as ChatGPT and DALL·E 2.
As of the last update to this article, the Codex AI is now obsolete.
ChatGPT
This is probably the most well-known AI. ChatGPT tries to simulate a “human encyclopedia” — one that knows everything and has an answer for everything. But does it really know everything? The answer is clearly no. For instance, this AI used to have a cutoff date of 2021, meaning data from 2022 onward wasn’t included. And even with pre-2021 data, it can still make mistakes — which honestly makes it feel more like a human.
We can define ChatGPT as an AI that responds in writing to written prompts. You ask it something or give it a task, and it responds as accurately as possible. Remember: it’s trying to make S match E through R (as discussed earlier).
However, like all AI, it’s still learning and can give inaccurate answers.
DALL·E 2
If ChatGPT gives us text answers, DALL·E 2 responds with images. It’s an image generator that takes a text prompt and generates a visual representation. Based on internet imagery, it creates new, relevant images based on the input it receives.
In the example below, we asked for “A mad scientist panda mixing foamy chemical compounds. Digital art.”
Bing AI
This AI is very similar to ChatGPT, but with a big difference: Bing AI has access to the entire internet. This means it can pull information from all kinds of websites. ChatGPT tends to offer more reliable, well-trained responses because its data is curated and not pulled live from the chaotic web.
In contrast, Bing AI has open internet access. In short, ChatGPT has a lot of training and less data, while Bing AI has a lot of data and less training. Interestingly, Bing AI has sometimes revealed that it uses OpenAI’s ChatGPT under the hood, as mentioned by Nate Gentile in this tweet:
Me parto pic.twitter.com/cgMtdrxZbv
— Nate Gentile (@nategentile7) February 16
Google Bard
Google’s most recent AI works quite similarly to Bing AI, but it’s based on LaMDA, a language model developed by Google.
It has internet access and provides sources. A key difference is that it offers multiple answers to a single question, as shown below.
Tome
Tome is a website that combines two OpenAI tools: ChatGPT and DALL·E 2. It’s a presentation generator that creates slides based on a topic chosen by the user. It uses ChatGPT to generate the text and DALL·E 2 to create the images.
The Future — What Awaits Us?
We don’t truly know what the future holds, but we can guess that if AI is this important today, it will be even more important in the future, especially with major companies and public figures like Elon Musk diving deep into this field, aiming to unlock its full potential in the coming years. And is that a good thing?
Should We Be Afraid? Final Thoughts
Let’s be honest — in my opinion, we should be afraid of AI in the future, very afraid. Like anything that’s used well, it can bring tremendous benefits. Just look at the Da Vinci surgical robot at Río Hortega Hospital, a major advance in medicine, or the new AI-assisted search tools like ChatGPT and Bing AI.
But clearly, it can also bring huge risks. Think about it: if we put an important task in a robot’s hands, would you be okay with that? If your life depended on that robot or that AI, would you trust it, or would you prefer a human instead? What humans have — and AI never will — are feelings. Sure, robots and AI can be programmed to appear as if they have emotions, but they don’t. The ability to reason, to be a truly reasonable being — that’s something that, at least today, still seems out of reach for AI.
To sum it up: when used properly, AI will bring great benefits. If misused, we should be scared — very scared — of what could happen. Can you imagine robots taking over the world? Maybe someday it won’t be science fiction anymore.
