Is Artificial Intelligence tech already relevant to our everyday life?

Lately, most Artificial Intelligence (AI) articles that dear Google has been suggesting me either explained (1) how predominant AI will become, (2) how it will steal many/most/all possible jobs or (3) how the singularity is around the corner (a.k.a Terminator for people born before 2000). As a researcher working in the domain that gave birth to a considerable share of these technologies (Natural Language Processing) and who knows a few of their inner workings, I think all three points are, for now at least, fairly exacerbated by people's imagination and fears, as well as short-term economic interests. My stance towards these futuristic scenarios is thus similar to the one I have regarding space exploration: it will undoubtedly happen someday, but I don't think I'll be travelling to the moon anytime soon.

Yet, nowadays’ AI tech can already achieve some impressive deeds worth knowing about. So, let's take a moment to review some of them related to generating photos, pictures, music, speech, discussions and video.*

For example, do you recognize anybody on this website? If you don't, that's normal: none of the people in the pictures actually exist. Detecting AI-generated pictures and photos has become a major concern, especially for artistic contests like the one in which such an AI-generated picture won the first prize.

In terms of music, I invite you to test recent automatic generators, which, as far as I can tell, are doing a fine job with "codified" music styles such as dancing ones. My creation "Eurac grooves" generated with AI Music and the simple prompt "a salsa music about a research institution called Eurac", left me so speechless and hilarious that I included it in my playlist (and you’ll probably hear me humming it if you come around).

AI’s capacity to generate discussions is nowadays well-known. But did you know that ChatGPT was the first ever chatbot to truly overcome the legendary decade-lasting "Turing test" for which humans have to detect if their interlocutor is a machine? Combining this capacity to discuss with the upgraded one to process and generate speech also gave birth to a new wave of enhanced spamming robocalls. I personally started receiving such calls some two months ago and I must shamefully admit that it first took me at least two minutes to understand that I was talking to a robot. Funnily enough, this idea is also used against scammers: meet "Daisy", a sympathetic grandma AI bot devised to waste scammers’ time and energy, thus reducing their capacity to scam real people.

As for videos, let me tell you about the ugly story that a poor accountant in Hong Kong went through (and would make a fine script for a movie). After wiring a hefty sum worth of 25 million dollars on an order given during an urgent conference call by his company’s chief financial officer, the unlucky accountant realized that none of the coworkers could remember the call, as if it had never happened. It turned out that everybody in that call except him was a scammer impersonating one of his colleagues behind automatically generated deepfake videos. Unfortunately for the poor sir, their meticulously prepared plan worked out just as planned.

AI technologies to generate other types of content (such as news articles, programming code or even scientific research) are also nothing to sneeze at and would be worth discussing. But I’ll leave that for future posts and sum up this post as weel as my current opinion on AI: while I don't think it will bring doom (or salvation?) upon us anytime soon, it is certainly already capable of impressive feats that I would have had a really hard time foreseeing some few years ago.

Last but not least, if someday you end up discussing on the phone with an adorable English grandma named Daisy, please send her all my love!

( * ) If you'd like to know more about how AI technologies relate to our work at the Institute for Applied Linguistics, feel free to check some of our past articles such as https://www.eurac.edu/en/blogs/connecting-the-dots/wie-gut-ubersetzt-chatgpt or https://www.eurac.edu/en/blogs/connecting-the-dots/tell-me-chatgpt-how-man-meters-can-t-you-jump.