)
)
You might have noticed that, as your friendly neighbourhood software developers, we’re on a bit of an Artificial Intelligence (AI) journey at 3 Sided Cube. And by bit, I mean, we are DEEP into our AI era, hard at work at HQ experimenting and tinkering with different applications to ensure we aren’t here to merely participate, but forever innovate.
I’m Joel Dolling, Project Manager extraordinaire at 3SC and here to contribute to your AI regularly scheduled programming. As a PM, I’m already seeing an uptake on certain AI applications, but today, I wanted to dive into a comical/concerning trend I’m seeing with “AI Washing”.
What is AI Washing?
The use of advanced tools is the difference between humans and animals. There is no other species that does it like us. From the seed drill that kicked off the agricultural revolution to the internet that gave us the information age. Usually, the use of a tool is absolutely unmistakable. For instance, you know if someone calls you using a phone because your phone rings.
Currently, we are sitting on a rocket that’s lifting off. AI promises to give us a revolution reminiscent of the internet. The thing is, AI is such an attractive marketing prospect that I’m dubious that everything that is marketed as AI is actually created by AI.
Right on theme, I turned to my steadfast and know-it-all new friend, ChatGPT to tell me what AI washing was and this is what I got back,
AI washing is the deceptive practice of overstating or misrepresenting the use of artificial intelligence in products or services to make them appear more advanced or innovative than they truly are, often for marketing purposes.
ChatGPT, The All Knowing Robot Overlord
AI is the buzzword du jour, so naturally some companies are hopping on the bandwagon to capitalise on the fascination and volume of searches surrounding AI, labelling their products or services as AI-driven or intelligent systems to appear cutting-edge, even when the technology is not truly innovative or intelligent.
This new wild west is in its infancy (and the gold rush is ON!) and it’s easy for consumers, investors, and the general public to be misled by overhyped AI technology. Creating unrealistic expectations about the capabilities of a particular product or the level of AI integration.
Forms of AI Washing to Watch out for:
Token AI: Adding minimal AI features or capabilities to a product or service without significantly enhancing its functionality or effectiveness.
Overhyping AI: Using AI terminology and buzzwords extensively in marketing materials, even if the AI components have a minor role in the overall offering.
Misleading claims: Making false or exaggerated claims about the capabilities or benefits of AI within a product or service.
Obscuring human involvement: Failing to disclose the significant human intervention required to make AI systems work effectively, giving the impression that AI operates autonomously.
Rebranding: Renaming existing technologies or practices as AI-driven, even when the core technology remains unchanged.
We’ve seen this kind of clever smoke and mirror marketing with “green washing”, but unlike oil giants making an arbitrary pledge to cut their carbon footprint (lol), AI washing is a little trickier to suss out. To avoid falling for AI washing, it’s essential for all of us as users and consumers to conduct due diligence and critically evaluate the claims made about AI in products and services.
Let’s put that critical thinking to use and go through a few examples to find out…
But is it AI though?
I’m not here to piss in your Weetabix, successful adoption of AI into a project can have a significant and marked effect with minimising laborious human intervention and future-proofing the scope. But not every so-called AI-powered product is what it claims to be. AI IS the future, we just like our AI served up with a side of honesty and genuine intent.
(Shameless plug: head over to our 3SC AI Labs to see very real examples of how AI is being used for good at your friendly neighbourhood tech agency)
The implementation of AI into an existing process is difficult. For instance, you cannot replace an entire human team with AI overnight. The writer’s strike in Hollywood is a great example of this, showing that human creativity is still worth paying for even though AI could complete the same tasks.
Before we swan dive into this one, we should first understand what AI is. AI is the implementation of machine learning into a tool so that the tool can update its own responses based on data inputs. Where I think the confusion arises with AI is that people assume it means that something is computerised. CNC machines, for instance, are not AI, they are very clever machines capable of taking onboard complex instructions, but they do not learn based on inputted data, so are not AI.
That is the difference between complex computerised machines and AI.
The AI washing verdict…GUILTY
LG’s AI Washing Machine
)
This is AI washing. Get it? AI washing because it’s an AI washing machine. You have to give it to the people at LG, they know how to have a good time. Lets recap the fundamentals of how AI and washing machines work:
AI works on data. This machine cannot harvest data (because it’s a washing machine)
Washing machines work fine anyway, and have done for some time now
Both of these points add up to the single, inconvenient truth behind the validity of an “AI washing machine” which is…
Even if a washing machine could access and interpret a large dataset that was relevant to washing, would it even make a difference?
I am personally of the opinion that washing machines fulfil the brief. We, as humans, don’t need them to get any better. How does AI ensure that my socks are clean?
Coming clean (pardon the pun), my brother has one of these, and once I tried to read the sticker on the front of it and couldn’t wrap my head around the amount of acronyms there were, so I’m thinking AI isn’t the only problem here. However, the experience of seeing one of these magnificent beasts in the wild made me think about the practicalities of the data that it could possibly interpret. I’ve come up with the following possible data points that could be interpreted by a dishwasher:
How large the load is (by weighing the contents of the drum)
How absorbent the load is (by comparing the soaked weight to the dry weight without any excess water
Any input the user makes, like the type of wash
If we were to assemble a table of these three variables you can have great control over the wash itself, things like time spent in the spin cycle, amount of detergent and how thorough the rinse is can all change based on these inputs, but here’s where we fall into the “AI marketing trap”. All of these inputs will have an output that is selected based on an algorithm, not based on any sort of intelligence.
To summarise, here’s why the AI washing machine isn’t necessary:
It is very likely to not contain AI
It definitely doesn’t use AI to be a materially better washing machine
Every single function of the AI washing machine is just as good in a traditional washing machine so what is the point?
Spotify: AI DJ
This one really gets to me.
In 2023, Spotify unveiled the ”AI DJ” which is a strange experience where a voice talks to you in between songs like a late-night commercial radio host. There’s a button to click if you don’t like the vibe of the music they play and they switch it up. The thing is, Spotify has been able to suggest content to you based on its algorithm and your listening history for years, maybe even a decade. There is nothing new about the songs the “AI DJ” plays because it seems to just use the Spotify algorithm, however, it does seem to be an AI-generated voice (the “DJ”) talking to you.
One thing that somewhat bothers me about the AI DJ is that it cannot be reasonably expected to be better at selecting music than the algorithm that already exists (and is definitely not AI). I think that this product was created more for the novelty of including AI, rather than AI being added because it adds material value to the proposition.
)
Conclusion: I’m prepared to concede on this one. The voice has to be generated by AI, but the songs it chooses are what I am dubious about. Try it regardless, nothing like a disembodied robot voice comin’ attcha in-between your favourite tunes.
Samsung Space Zoom Moon Pictures
Have you ever suffered the indignity of trying to take a picture of the sky at night with your phone? I imagine so. The next morning you look back at the picture and it could be mistaken for a hole in a wheely bin, taken from the inside. Well, Samsung has a fix for you. It’s AI powered lies.
)
Enter the Samsung Space Zoom. The concept is simple: “images of the moon that aren’t rubbish”. The implementation is much more complex, you just have to develop a better camera than has ever been seen on a mobile phone which can take good photos of the night sky, right? Wrong!
Instead of developing that camera, Samsung went down the software approach, and they made the moon zoom feature an AI copy. Here’s how it works:
Point your camera at the moon
Zoom in
The phone recognises the white sky blob as the moon and superimposes a good image of the moon on to it
I’m sure you have your own feelings about this, you might think it’s fine and you may not. I’m not here to judge, but this is a rare case of AI being used but deliberately concealed, rather than being used as a marketing tool. Props for that to Samsung I suppose, but maybe come clean about it in future.
Amazon “AI Supermarkets”
Imagine you do your normal weekly food shop but all you have to do is scan an entry QR code on your phone, then just pick stuff up from the shelves and leave. No tills, no queue, nothing like that. You still pay, but only because AI monitors the CCTV footage and can tell what you’ve picked up and charges you for that. What a utopia that would be. Well, it exists! Amazon has created this shop!
That’s the story anyway. What they failed to mention is that there’s a team of 1000 workers that check that the AI has got it right. It turns out that what Amazon has created here is 1000 perfect candidates for “The Price Is Right”. Not that impressive then.
All in all, it’s a cool concept for sure, but the ethos here that should be remembered is if AI cannot be held accountable, it should not make consequential decisions. Amazon has clearly realised this, but still said it was AI anyway. Cheers, Jeff.
)
As you can see, these examples are little more than non-AI-related tech and clever marketing, but while you’re here…let’s talk about some seriously cool applications of AI doing some undeniable good around the world.
Kick arse and for-good AI
Amazon Rainforest Deforestation Monitoring
At 3SC, we are passionate about stopping deforestation, so much so we built a product to help fight it.
AI has entered the chat in this fight now, with a team-up between Microsoft and Amazon who have combined to create a program where you input aerial photographs taken on different dates and it can work out where deforestation has happened for you.
Conclusion: This is a fantastic innovation and undeniable application of AI in the deforestation movement and could revolutionise the way that monitoring happens, which is manually currently, relying on locals to go and check areas they know. This removes the people from the equation, using satellite imagery and a computer program instead for instant and real-time results.
)
Image of deforestation in Brazil taken using the GFW partner’s satellite, Planet Labs.
AI-Assisted First Aid
Another Way AI is contributing to a better world is in the acute first aid space, where AI takes the fast-paced information gathering and output. Imagine watching Casualty (when it was on) and seeing someone getting wheeled from the ambulance to the hospital without the person trying to relay everything they could remember on the journey because the hospital team already knows everything about the patient and the situation thanks to AI.
Conclusion: This is something that we at 3 Sided Cube love to see. I wonder if one day, information input to our first aid apps could integrate with such a system. What a world that would be!
AI Disaster response
The number of disasters is projected to reach 560 a year – or 1.5 disasters a day – by 2030.
We’re all too familiar with waking up to news of the latest horrible disaster to be unleashed on humanity. The loss of life is an awful eventuality of this reality. But AI can help with that!
xView2 has been deployed in earthquake wreckage, wildfires, and flooding to successfully help workers on the ground be able to find areas that were damaged that they were unaware of. Every minute counts in a disaster and delayed arrival of search and rescue happens expedited and targeted with the help of xView2. Turkey’s Disaster and Emergency Management Presidency, the World Bank, the International Federation of the Red Cross, and the United Nations World Food Programme have all used the platform in response to the 2023 earthquake in Turkey.
AI is used by the algorithms employing a technique similar to object recognition, called “semantic segmentation,” which evaluates each individual pixel from 120 miles up in space of an image and its relationship to adjacent pixels to draw conclusions where to concentrate search and rescue efforts.
Conclusion:
The code is open source and the program, free to use. This incredible piece of technology is out there to save lives and there’s no one sitting behind a giant desk, stroking a bald cat and counting their billions with each disaster that rolls in – it’s just out there doing good! It seems future disaster response and recovery efforts should always include xView2.
And that’s a wrap!
Although it’s fun to poke at questionable attempts to bring AI-related products to market, I think there’s a bigger question about when it is appropriate to use AI in product development. For this section, I am taking from this excellent article which suggests three questions to ask to analyse whether the implementation of AI is necessary, warranted or ethical.
The questions are as follows:
Will the proposed AI use most likely be no worse than current reasonable human analysis and decision-making?
Does the proposed AI use reflect reasonable efforts to eliminate inaccuracy and bias?
Is there a reasonable chance that if the AI use is allowed to evolve, it will become fairer and more accurate than human efforts?
In the case of bandwagon-driven quasi-AI products like those mentioned earlier, I hope the wave crests soon. In my opinion, they generally fail question 1 for being worse than human decision-making. What I hope for soon, is products being developed with substantial AI influence ethically, rather than shoehorning AI into the value proposition of a product just because it’ll sell more units.
Time will only tell – watch this space for all our AI experiments and endeavours!
Published on 6 February 2024, last updated on 6 February 2024
)
)
)