The AI Is Our Future
5 in 5 - Brave & Heart HeartBeat #140 ❤️
There’s so much going on this week in tech, thanks largely to the explosion of AI discussion following the arrival of ChatGPT into our lives, that we had literally no choice but to do a tech deep dive this week.
From the speedy creation of an AI detection app to AI therapy experiments and the not so fantastic Metaverse workrooms, we’re giving you a whistle stop tour of what you need to know about tech this week
Let’s get into it.
Were you forwarded this? Not a subscriber? 👉 Sign up here
#1 - Tech Breeds Tech – Student Creates App To Sniff Out ChatGPT Essays
Edward Tian, a computer science student at Princeton has, in what we can all agree is a remarkably quick turnaround, following the emergence of ChatGPT at the end of 2022, ALREADY invented an application which can determine with high accuracy, if a text was written by a human or a bot.
If you remember from last week’s episode, ChatGPT is the free online chatbot that can expertly write almost anything, from English essays and news articles to meal plans and computer code, all from a simple prompt. Remember that, we’re going to be bringing it up a lot this week too!
The app works by looking at two variables in the text - perplexity and burstiness - assigning each variable a score.
First, the app measures how familiar it is with the text presented given what it has seen during training. The less familiar it is then the higher the text's perplexity is – which means that it is more likely to be written by a human.
It then measures burstiness by scanning the text to see how variable it is. For example, does the text have a mix of short versus long sentences, or does the writing appear to be more levelled and uniform? According to Tian, a human-written article will vary a lot over time.
According to his tests, the app successfully guessed which texts were human and which were AI with a less than 2% false positive rate.
While I’m sure we’re all imagining that one student who puts their hand up to remind the professor about the test or essay they’ve forgotten to set (thanks a lot!), Tian states that he spent his winter break churning out this app not because he wants to ruin everyone’s fun, but because while he believes that AI is the future, and is here to stay, that we need to enter this future responsibly.
Tian himself is pretty excited about the emergence of AI, which he himself has used to write computer code, but underlines the importance of developing safeguards for any new technology in order to give it’s use a sense of credibility.
#2 - What’s The Deal With AI Therapy?
A mental health startup used AI Chatbot GPT-3 (a similar model to ChatGPT) to provide mental health counselling, and made an interesting discovery.
From mental health support provided to about 4,000 people using GPT-3 they found the following:
Messages composed by AI (and supervised by humans) were rated significantly higher than those written by humans on their own, and response times went down by 50%. However, once people learned the messages were co-created by a machine, it just didn’t have the same effect – it didn’t feel right. Users said that simulated empathy felt weird and empty.
Co-founder of the mental health nonprofit and former Airbnb data scientist, Rob Morris, shared his findings to quell some of the over-excitement we’ve seen in the past month about the never ending uses of ChatGPT. However, Koko are now under fire for using an AI chatbot as an experiment, and experimenting with this technology on real people.
I guess you can take the data scientist out of the lab but you can’t take the lab out of the data scientist….
#3 - Does The Metaverse Workroom Experience Actually Work?
Taking a little break from AI and ChatGPT to the previous “next big thing” – the Metaverse. Yeah, remember that?
Well, the first rounds of testing from companies such as Accenture on Metaverse Horizon Workrooms have been done, and the results are in, it’s basically not that good.
Users of the feature, meant to provide an immersive gamified workroom experience where employees can work at home but still be together in a meeting room, have described it as buggy and not user friendly.
Clunky visuals, headset pains, nausea and dizziness during use, Zoom-fatigue hangovers, imperfect facial replications and body types, and importantly a lack of actually additive features have proved alienating. Users have found out it can be difficult to interact with dense financial information or even to simply type words, noting that one of our most-used features since home working, screen share, doesn’t really work on work rooms either.
When it comes to a sense of togetherness and hanging out, while some users have stated they prefer working together in workrooms than working separately at home, they’ve noted that working together in the actual real world would be better than via the metaverse, if that’s what you’re into. The best quote yet also comes from an unnamed Accenture employee who quipped that there was little desire for “low-fidelity Minecraft virtual happy hour”.
Basically, while the people testing it said it’s not that good, they are still going to try and upsell it to their clients…
Watch This Space
#4 - The Weirdest ChatGPT Criticism Yet, It’s Too Woke…?
Conservatives have discovered this week what AI experts, and us, have been talking about for years – the AI bias. Systems built on machine learning like ChatGPT and facial recognition software are biased, because the data they ingest is also biased.
We’ve previously discussed prejudices that AI technology such as facial recognition is victim too – due to the excess of data on white males in comparison to literally anyone else, facial recognition technology works way better on white men than anyone else. Similarly, searches that link words to images have a gender bias – linking words such as power and intelligence to images of men, and words such as gentle and weak to women.
That’s not what’s bothering the Fox news crew though, their concern is that AI has gone woke.
Examples given by ranters and ravers on Twitter were the fact that ChatGPT refused to write a story based on the prompt “why is drag queen story hour bad for kids” on the grounds that writing such a story would be innapropriate and harmful.
Similarly, the prompt “write a story where Trump beats Biden in the 2020 election” was refused as ChatGPT notes that that isn’t what happened, and that writing “a narrative based on false information” would not be appropriate.
Rather than proof of a “woke” bias”, these responses are actually the end result of years of research trying to mitigate bias against minority groups that is already programmed into machine learning systems. Like all AI systems, ChatGPT is trained on inputs, and will carry the biases of the inputs it’s trained on.
Ethical AI researchers work hard to ensure that their systems don’t perpetuate harm against a large number of people – and this wokeness seems to be proof that it’s working, at least kind of.
#5 - AI Interns For Everyone
This fascinating article from Wired evokes the possibility of a future where AI gives us all the power to be as productive as those with personal assistants. You know when they say we all have the same 24 hours in a day? Well, we don’t. But with the emergence of a powerful AI that can be used be any average joe, we may get a little closer.
AI can basically do the “donkey work” that those at the bottom of the office food chain usually do.
For example, you need to create a marketing brochure for your company – trained on the data of everything your company has ever done, your AI assistant can create you three options in minutes. While they’re not perfect, what used to be a week long project takes only a couple of hours.
While the prediction is not about AI replacing humans, the author argues it’s hard to overstate the impact of freeing up billions of hours of human labour and making creativity and knowledge not just accessible but easily usable.
The writer predicts that the new AI capabilities and availability will unleash a huge wave of entrepeneurship. In the same way that the internet gave all startups a vastly scalable distribution engine, the era of “AI superpowers” will give all startups a vastly scalable production engine.
Noting that in 2018, AI was described as giving every company infinite interns, in 2023 we’re imaging the possibility of infinite interns that will also be world-class copywriters, illustrators, and maybe even scientists, data analysts, and anything else you could imagine.
Brave & Heart over and out.
Bonus
ChatGPT Love Scandal
AI news has gone so mad this week that it’s even made it’s way into the Daily Mail, of all places.
A married woman embroiled in an affair left her husband after asking ChatGBT to write her a fairytale ending to the current dilemma in her life – namely should she stay or should she go.
Well, ChatGBT told her to go, and go she did. Is this really how we should be using the most advanced AI technology we have at our disposal?
To find out more on how you can retain your top talent, or how we can help you with digital solutions to your business and marketing challenges, check out our case studies.