Here Is How You Get Friendly With A.I. Before It Gets To The Office

If you ask a random person whether they use artificial intelligence in their everyday lives, the answer will likely be a firm negative. However, this answer is incorrect: most of us use it a good number of times every day. A.I. is everywhere around us, it just doesn’t look like the mental picture we have, which usually resembles T-1000 from Terminator 2: Judgement Day or the Borg Queen from the Star Trek series. Ugly, evil, and almost omnipotent.

All images in this article were generated by Midjourney, a text-to-art algorithm. This was its take on "The Borg Queen and T-1000 from Terminator 2 in one picture"

Most of us use A.I. all the time: when navigating through the town with an app, when opening a social media platform and scrolling through its recommendations, browsing on an online marketplace, interacting with a chatbot or talking to Siri/Alexa/Google. There are algorithms everywhere, we just don’t recognise them as artificial intelligence, although this is what they are.?

However, things will not stop at this point, and in the coming decades, we will have to form a deeper relationship with A.I. as it will become a prominent part of our daily lives on a more active level. In the healthcare setting, for example, we can be assured that A.I. will become a member of the care team - so we better figure out how to deal with it.

A.I. will take over repetitive and data-based tasks and will transform your work

It is easy to determine how not to relate to this upcoming new era. Getting paralysed with desperation “A.I. will take my job and there is nothing I can do about it” (potentially with mental pictures of ending up under a bridge) is not the way.?

Ok, so how to relate? First, make it clear: yes, A.I. will take over many, many tasks from humans. And it will execute them faster, better, and cheaper. Tasks that are

• data-based
• repetitive

will be taken over by algorithms, whatever your line of work is.?

This process will not be materially different from what we experienced with multiple technological revolutions centuries or decades ago. Computers took over a good number of tasks from humans - accountants don’t have to do all the calculus for a giant firm on a notebook with a pencil anymore, but computers did not take the work of accountants, they lent advanced computational skills to them. Computers also created new jobs, jobs we never knew in 1980 would exist in a few decades.??

A similar shift is expected with algorithms and machines becoming better at certain tasks. A nurse might not need to do the blood draw in the future - machines will do it better -, or try and move heavy patients alone while changing their sheets. But this will not eliminate the need for nurses, it will transform what and how they do. They will probably have more time to empathically respond to patients, more time to talk about their cases and help them understand how to manage their conditions.?

Learn to deal with A.I. now, in a non-commital way

To prepare for the future, the best we can do is to familiarise ourselves with how artificial intelligence works - outside of our professional lives. This, again, will be a cultural transformation, and it is much better to learn it before it gets to the office.?

Midjourney, the text-to-art image algorithm is a good example. Granted, you need to have (or register) a discord account to use it, but after that, you are free to play with it for a while, get a grasp of how it “thinks”, and what is the best way to interact with it to get the result you imagined.?

I wanted to see if Midjourney can generate a picture of a futuristic hospital, so I told it what I typically tell our graphic designers with a task like that: I’d like a picture and the topic should be “futuristic, hospital, doctors, AI”.

Well, Midjourney certainly needs more details than Adam, our lead designer, as this is what I got at the first go:

Apart from the interesting fact that futuristic doctors seem to be exclusively caucasian females in Midjourney’s books, this was quite far from the results I wanted to see.

So I went on and defined the task better: “A hospital environment with patients, beds, healthcare devices, where doctors, patients and robots work together in a futuristic hospital setting.” One of the 4 results seemed to get closer to my concept, so asked Midjourney to refine the fourth picture, and we ended up with this:?

As this was still quite far from what I originally had in my mind - an isometric scene of a futuristic hospital, full of patients, devices, human and non-human personnel, something like this or this - I tried to define the task even better, asking Midjourney to imagine an “isometric image of a futuristic hospital building suit with patients in beds, machines, healthcare devices, human doctors and robots working together with AI”.?

Results this time got further from my mental picture, they were neither?isometric nor grasped the buzzing futuristic hospital scene I wanted to see:?

Ok, one last go before you get super bored with my imaginary science fiction hospital. I tried to define the task much better, by breaking my initial mental picture into shorter and simpler sentences and stressing the missing details.?

So it went: “buzzing hospital emergency unit setting with many people in the picture. There are several patients in hospital beds. Lots of medical machines and devices. Several human doctors and human nurses. Also working with medical robots and A.I. Futuristic setting.”

Looks like this did not help a lot, while each of the four images incorporated some new details, none of them incorporated all of it. There are many hospital beds in the first one, a good number of healthcare workers in the second one, a robot in the third, and a sort-of-buzzing emergency unit in the last one.?

I’m not sure how long I have to keep trying to redefine the task until I eventually get there, I suppose I will exhaust the free limit of 25 tries before I succeed in my quest. But all these queries helped me understand slightly better how this particular algorithm thinks and adjust my thinking to it.

I will take a break from Midjourney for now and will get back in a little while, trying to figure out how to get to the result I want to see.?

Is it good or useful? - you might ask. Well, yes and no. No in a way that we obviously can’t use Midjourney to generate images for publication - apart from this article. Also no in a way that after a number of tries - not all are documented above - I didn’t even get close to the picture I wanted to see.

On the other hand, it was useful. It helped me think about my thinking process. It helped me think about the algorithm’s thinking. It was interesting to see how it gets to a visual answer and what kinds of answers it provides.?

Be curious and never stop playing with new technologies

The above examples with Midjourney show how un-straightforward it is to use a text-to-art application - said to be the best at the moment - to make it generate the picture you want. However, it also shows that it can come up with many things, and some of these might surprise you. This is, what the algorithm thinks an overworked nurse looks like:?

Quite some pieces of socio-documenting art, aren’t they?

So I urge you to stay curious and open-minded, and grab the opportunities to play around with algorithms, should it be Midjourney, the A.I. in your favourite game, or the chatbots you interact with. Get as comfortable with them as possible and form a “relationship”, as you will need it in a few years and decades. And never think you will be left behind, instead focus on how it will enhance what you are capable of doing.?

Although I was dead-set on pushing out a header image of the algorithm, my first attempts on a "16:9 ratio image of a human doctor and an A.I. doctor shaking hands in the hospital" ended up with quite disturbing results (and a square image format - note: you need to use aspect ratio or the --ar command). Even if we skip the confusing faces, it is slightly unsettling how Midjourney seemed to take the "shaking hands" part a little too literally.?

So, to be on the safe side, I stuck with a simple command of "doctor in a futuristic hospital with a big AI terminal --ar 16:9", this is the header image of this article. Pretty not bad, right?

All in all, good job, algorithm, it was fun to work with you!?