This week we saw some exciting sneak previews of AI developments in the works, as well as more open-source product releases from Meta. And just when you think the excitement about AI is a little overhyped, investment and development seem to push on unabated.
Our roundup this week feels a little like the good, the bad, and the ugly, as we were reminded that there are some real ethical and environmental issues behind the cool AI tech we’re so excited about.
AI vs the environment – It’s not looking good.
Training AI models and getting ChatGPT to churn out your mid-term paper takes a lot of computer processing power. It turns out that the huge amount of electricity that data centers are using isn’t doing us any favors as far as our carbon emissions problem.
The servers running the AI models also need cooling and the most efficient way to do that is with water. And we’re talking serious amounts of water. Google alone consumed around 4.3 billion gallons of water. It looks like AI’s impact on the environment is going to get worse before it gets better.
The AI doctor will see you now
A lot of the AI medical chat has been around Google’s Med-Palm’s ability to answer patients’ questions and help with diagnoses. But this week we reported on how AI is being used to detect breast cancer.
Doctors have also implanted chips into a patient’s brain and used AI to retrain the neural networks in his brain and spinal cord.
Doctors hope this experimental AI brain rewiring will help treat paralyzed, blind, and deaf people. Once they’ve got that right, it would be nice if they could find a way to download a language or new skill straight into our brains.
AI investors: “Take my money!”
There is no slowing down in the amount of money being thrown at AI companies. The usual suspects at the top of the S&P500 tech companies delivered impressive Q2 returns and investors keep sending investment cash their way.
We’re not surprised when we read that the likes of Meta, Nvidia, or Microsoft attribute a lot of their results to AI but it turns out that AI is being mentioned across virtually all sectors in earnings reports.
And it looks like a lot of folks working at those companies will be changing careers in the next few years because of AI.
As the big guns at the top of the tech pile keep getting bigger there’s been the inevitable talk of a potential “AI bubble”.
We took a look at whether the pace that these tech companies are growing is sustainable or not. Will they retain the flexibility the fast-paced AI race seems to require or become constrained by their success?
Got AI? China says you’ll need a permit for that.
As the EU and US legislators quibble about what laws should be put in place to regulate AI, China just went ahead and did it. The new laws regulating generative AI in China go into force on 15 August. One immediate consequence is that Apple has pulled a whole bunch of apps from its App Store in China.
The need for regulation and oversight is highlighted by the ugly side of what it takes to sanitize AI for our consumption. It turns out that AI moderators in Kenya have had to look at some horrible stuff to make sure that we don’t have to.
AI is dangerous, let’s make it more dangerous
The potential risks associated with bad guys using AI have been heightened by how good deep fake voices have become. Think you could spot an AI voice? It turns out that over 25% of participants were tricked by these voices in a recent study.
As the dangers of benign AI applications are becoming more evident, the CEO of defense software company Palantir says the US should be making AI weapons. Yes, they will be dangerous, they’re meant to be. While the US wonders about whether or not they should, its adversaries are probably less reticent.
To which James Cameron says, “Uh, guys. Did you not watch my Terminator movies?”
Exciting AI releases
OpenAI has been working on DALL-E 3 and gave 400 people access to test their new image generator on its test server. Sadly we weren’t one of those people but the leaked images look awesome!
GPT-5 hasn’t been released yet and Sam Altman says they’re in no hurry to work on it. But we don’t believe him because OpenAI just filed a patent application for GPT-5. With the Turing test fast disappearing in the AI rearview mirror, there’s talk that GPT-5 could deliver AGI.
Meta keeps dishing its AI models out for free and did it again this week with the release of AudioCraft. The software is made up of 3 models that generate music and sound effects from text prompts and compresses the audio files up to 10 times smaller than MP3 codecs can. The samples on its blog were impressive.
In other news…
Android users no longer have to put up with the condescending looks from Apple fanboys. OpenAI finally released ChatGPT for Android. The Google Play store shows that it’s had more than 5 million downloads already.
Nvidia released the next iteration of its text-to-image personalization model called Perfusion. It works better than a lot of other image personalization models and at 100kb it’s small enough to fit on a floppy disk. Don’t pretend like you don’t remember what that is.
Probably one of the most expensive domain names changed hands this week. Previously if you type ai.com into your browser it took you to OpenAI’s website. Now it redirects to Elon Musk’s x.ai URL. Not sure how he wangled that. It seems like besides making electric vehicles, boring huge tunnels, and sending rockets into space, Musk still has time for AI vanity projects.
And finally, let’s face it. You need all the help you can get. Tinder is testing an AI tool that spices up your bio and chooses the best photos of you so that you actually might have a shot.
And that’s a wrap. So, what can we expect from the AI industry next week? To be honest, we don’t know. Heck, even the experts don’t know. And we love that! But we’ll be here to keep you updated on the latest AI news as it happens.