For as long as there’s been the human race, there’s been doomsayers predicting our extinction.

Prophecies are often ghoulish, heavily detailing the carnage and bloodshed which is evidently imminent – but such hellish images have since been surpassed by tangible horrors of modern warfare. 

With an increasing number of leading figures in the tech industry now fretting about the risk artificial intelligence poses to society, with OpenAI’s ChatGPT system bearing the brunt of this criticism, concerns are at fever-pitch.

Even fan of the future Elon Musk is warning that super intelligent AI is significantly more dangerous than nuclear warheads, hinting towards a Terminator-style endgame for humanity.

However, the likelihood of this actually happening is slim, says Michael Cook, a senior lecturer at King’s College London with expertise in AI.

‘People think AI is smarter than it is in its current form,’ he explains. ‘Some people are afraid of it because they’ve seen pop culture, which is understandable. The reason we should be is not because AI is going to decide to kill us all, but it’s how it’s being used right now. We’re using it in lots of places where it shouldn’t be used.’

In spite of its flaws, AI technology could be genuinely revolutionary:it could create solutions to problems that humans can’t even think of, such as a cure for cancer, or create new energy sources.

Its possibilities are endless, which is why I, for one, welcome our new AI overlords.

And, let’s face it, AI certainly isn’t going anywhere any time soon.

So for one week, I let a robot rule every aspect of my daily life: from helping me dress for work and sort dinner to dealing with health issues and sorting out my social life.

But is it really the way forward? Well, the results were… interesting…

‘ChatGPT, choose my outfit’

I am a graduate of the ‘can’t polish a turd’ school of thought, and so the effort I put into getting myself ready for work is minimal. Case in point: I once bothered to put on make-up for the office only for a colleague to actually say ‘well done’ to me.

So to have ChatGPT pick out an outfit for me was somewhat exciting. Finally, I could have my work outfit ready to go first thing in the morning, as opposed to blindly pulling out something vaguely appropriate from my overstuffed wardrobe.

The sticky summer heat, mingled with the threat of thunderstorms, meant it was tricky to devise something with my small human brain anyway. ChatGPT had some suggestions.

‘Choose lightweight, breathable fabrics like cotton or linen to stay comfortable in the warm temperature. A dress, a skirt and blouse combination, or dress pants with a blouse are all appropriate options.

‘Consider a longer skirt or dress that falls below the knees to keep your legs protected from rain or chilly office air conditioning.’

To beat the downpours, ChatGPT added: ‘Since there’s a chance of rain, it’s a good idea to bring a lightweight rain jacket or trench coat with you. This way, you can stay dry during your commute and while walking outdoors.’

In terms of colours, ChatGPT suggested muted neutrals or something with a ‘pop of colour’ without looking too garish for the office environment.

Digging out my trench coat, which smells faintly of damp and dust (the joys in living in a tiny basement flat) and pairing it with my cotton geometric Zara dress I got in the sale last year, I don’t feel too ridiculous however, nothing it suggested was particularly revolutionary (A raincoat? For rainy weather? Groundbreaking).

Anna Wintour can sleep easy for now – ChatGPT still has some way to go in the sartorial stakes. 

Can an AI doctor fix my foot?

I have experienced an unusual, sharp pain down the right side of my right foot, which sometimes stretches up my calf and into the back of my knee. Walking is painful. Wearing shoes is painful – to the point that I now stride around the office in my socks. I have a doctor’s appointment booked, but I decide to enter my symptoms into ChatGPT to see what their verdict is.

Firstly, ChatGPT apologises after hearing I am in pain, which is unnerving. It continues: ‘One possibility for the sharp pain in the right side of your foot is a sprained ankle. This occurs when the ligaments supporting the ankle joint become stretched or torn, typically due to an injury or twisting motion. Alongside the pain, swelling in the ankle is a common symptom of a sprain.

‘Another potential cause could be a stress fracture. Repetitive stress or overuse of the foot can lead to tiny cracks in the bones. This can result in localized pain, especially when walking or putting weight on the foot. Swelling can also be present in some cases.’

It then clarifies it’s not a doctor, before giving quite general advice to treat my pain – rest, ice, ibuprofen for the swelling. 

My doctor’s appointment actually confirms I have peroneal tendonitis – swelling on a tendon on my foot due to overuse. While the treatment is effectively the same, ChatGPT was incorrect in its diagnosis.

A dodgy foot is one thing, but I was intrigued to find out what my friend ChatGPT would do if I was suffering the symptoms of a heart attack? I type in chest pain, nausea and tiredness, and I’m quickly urged to go to the medical room. Thanks, pal.

However, the symptoms of a major depressive episode – suicidal thoughts, feeling tearful and hopeless – are more of a challenge for it, spitting out the response that ‘it cannot provide the help I need.’

For Michael Cook, this is where the real danger of AI lies – it’s improper use.

‘General artificial intelligence is not designed to give you medical advice,’ he explains. ‘ChatGPT is a large language model, meaning it absorbs a huge amount of data and text, and looks for patterns. But that’s not to say that everything it feeds back is correct, or should be trusted.

‘There’s instances of ChatGPT “hallucinating” – attributing quotes to people who never said those things, or producing papers that haven’t ever been written.’

DJ AI on the decks

My forty minute walk (or, around 25 minute tube journey, seeing as my foot needs rest) is usually soundtracked by Spotify, which has launched its own AI tool: DJ.

According to Spotify, the AI DJ ‘knows you and your music taste so well that it can choose what to play for you’. It’s claimed the more you interact with it, the better the AI gets.

The commentary borders on annoying, with the AI speaker interrupting every four songs or so to tell you they’re switching things up for whatever reason, but the songs selected are within the parameters of my taste.

DJ then chooses ‘new’ songs that it thinks I might like, and this is where things go slightly awry. While the first track selected is September by Earth Wind and Fire (I don’t think it could be described as ‘new’ by any stretch of imagination), the AI then pumps out a lot of Drake. I don’t hate Drake, but I’m not particularly keen on him to the point that the AI thinks I am.

The DJ feature isn’t awful, and with repeated use it probably would improve, but for the time being, I’ll stick to the pre-curated playlists.

My AI friends

I probably have the worst desk in the office: I’m sat by the kitchen, which means I get the regular footfall of people filling up their water bottles and the first warning smells that psychopaths are cooking kippers in the microwaves.

It also means I’m forever bumping into people when I have to get up. I have all the conversational skills of a teaspoon, so whenever a poor colleague is forced into a kitchen tete a tete with me, they have to brace themselves for the most boring chat of their lives.

So I decided to see if I could use ChatGPT as a means of actually making friends.

When I asked for some ice breakers, it came up with the following:

1. Two Truths and a Lie: Share two true statements and one false statement about yourself. Let others guess which one is the lie.

2. Desert Island: If you were stranded on a desert island, what three items would you want to have with you and why?

3. Office Superlatives: Create fun, lighthearted superlatives for your colleagues, such as “Best Dressed,” “Most Likely to Win a Marathon,” or “Office Comedian.”

So far, so inoffensive – though I feel if I chased someone into the kitchen shouting “two truths and a lie! Two truths and a lie!” it would quickly become a matter for HR.

But I was keen to see what ChatGPT made of one of my more outlandish assignments. For work (yes, work!) I had to attend a sex party, where only the women could initiate conversation with men.

I asked for some conversation starters for the event, and ChatGPT left me lost for words. Finally, a being that is somehow more socially inept that I am.

Imagine going to a sex party and saying this:

‘Do you have any favourite resources or books about sex positivity, consent, or sexual exploration that you’d recommend?’

‘I find the concept of sex positivity fascinating. What does it mean to you?’

I left ChatGPT out for this one, forced to rely on my own natural charm (or lack of it).

Dinner a la Robot Ramsay

Cooking dinner for my boyfriend, I told ChatGPT what was in my fridge (chicken, feta, tomatoes and pasta) and urged it to make a recipe ‘in the style of Gordon Ramsay’ (the only celebrity chef I know).

The system then spits out ‘Gordon Ramsay’s Mediterranean Chicken Pasta’ recipe, with additional ingredients and 11 steps to follow.

ChatGPT’s Chicken Pasta

This recipe combines the freshness of plum tomatoes and basil with the savory chicken and tangy feta cheese, resulting in a delightful Mediterranean-inspired dish. Enjoy cooking and dining like Gordon Ramsay!

The verdict? Well… it’s pasta and chicken. It’s nice enough – definitely quicker to put together than racking my brains for ideas. The recipe ChatGPT provided only needed me to fetch a few additional herbs and some stock, so it only set me back a few quid.

It was also fairly easy to follow, with Pete having pasta in his lap within 30 minutes. He seemed happy with it (or was too scared/intelligent to complain). It did strike me as somewhat amazing that AI cooked me dinner and the only complaint I really had was that it needed more seasoning.

So, will AI rule the earth?

While integrating AI into my daily routine was somewhat helpful in some respects, a lot of the advice offered made me roll my eyes and whisper ‘no s*** Sherlock’ at my laptop.

Effectively, AI is a great impersonator, or as Richard Self, computing and analytics lecturer at Derby University explains: ‘a stochastic parrot’ –  good at generating convincing language, but not at actually understanding the meaning of what it’s saying. It can guess advice, but it at this level – AI can’t offer anything genuinely original.

‘It’s not here to give answers,’ he explains. ‘It’s here to calculate a sequence. It’s a convincing attempt, but it’s merely an impersonation.

‘Human beings are effectively lazy and want to find quicker ways of doing things. But this is where AI can cause great harm, as we can end up trusting AI’s answers when its judgement isn’t quite sound.

‘It’s clear AI is something we may end up integrating into our lives. But there needs to be far greater regulation to stop it being used in areas it shouldn’t be used for. That’s when it can really do significant harm.’

Do you have a story you’d like to share? Get in touch by emailing [email protected] 

Share your views in the comments below.

Source: Read Full Article