Welcome to this week’s roundup of AI news written by humans for humans.
This week AI and VR made us plug in and disconnect.
OpenAI says it can’t stop ChatGPT from lying.
And AI bots are taking over the internet.
Let’s dig in.
Virtual reality, real isolation
Advanced communication technology has made humanity more connected than ever. Paradoxically, advancements in AI and VR have seen users isolate themselves in their own hyper-personalized virtual realities.
With VR I can experience a “reality” where everything is exactly as I want. Can that be a bad thing?
Sam took an in-depth look at how VR is making people more insular and how society can mitigate isolation in a virtually connected world.
ChatGPT lies. Deal with it.
Privacy advocacy group noyb (none of your business) has sued OpenAI because ChatGPT keeps making up false info about people. OpenAI says that keeping its models honest is a “work in progress” and that, for now, the problem can’t be fixed.
We know transcription tools aren’t perfect. When a speaker says “black” and the tool transcribes this as “block”, we put it down to an honest mistake.
When AI transcription tools like OpenAI’s Whisper make mistakes, things get weird. During long pauses in spoken audio, Whisper hallucinates to fill the gaps and it often adds some disturbing content to otherwise innocuous conversations.
OpenAI’s approach to wholesale data scraping has made it the target of criticism and lawsuits.
The company maintains its stance on “fair use” but has outlined plans for a “Media Manager” platform for creators to allow creators more control over how their content is used.
‘Hey guys, sorry we stole your content to train our models. We’re going to keep doing it until 2025. But we promise we’ll have something in place then so you can decide how we use the stuff you created.’
Right pic.twitter.com/Zbuw5JcN79
— Bojan Tunguz (@tunguz) May 5, 2024
Poached AI
Apple’s AI strategy sounds like it was crafted by James Bond scriptwriters rather than board execs.
The company has been poaching AI talent from companies like Google and setting them up in a secret AI lab in Zurich. What are they working on?
Microsoft did a talent grab of its own earlier this year when it took on most of Inflection’s engineers and its CEO Mustafa Suleyman. Suleyman is reportedly heading Microsoft’s project to build a 500B parameter LLM called MAI-1.
Once released, MAI-1 will be the largest AI model Microsoft has ever built.
AI on auto-pilot
AI is automating loads of processes that we previously had to do manually. And that’s not always a good thing.
A Princeton-led team says that AI poses risks to scientific integrity. Machine learning processing of large data has supercharged the publishing of scientific research. The problem is that ML-driven research is often impossible to replicate which is essential if it’s to be trusted.
Have you heard of the “Dead Internet” theory? The short version claims that the majority of web activity is AI-generated with humans increasingly taking a back seat.
How much web traffic do AI bots account for? A new report says it’s pretty much 50/50 and a lot of the bot activity is sketchy.
Is automating weapons of war using AI a good thing or not?
The debate continues while that reality unfolds. US Air Force Secretary Frank Kendall took a flight in an AI-piloted F16 fighter jet.
It sounds impressive, but should we be handing over the keys of an F16 to AI while the hallucination problem remains unsolved?
Talking AI
We’ve been getting to speak to some very interesting people in the AI and blockchain space. The symbiosis between these technologies is fueling some exciting developments.
Eugene spoke with Cinderella Amar, a Web4, AI, and Blockchain evangelist. Amar’s Glass Ventures VC fund is involved in some very interesting Web4 projects and her passion for the tech is infectious.
Chad Sanderson is the CEO and founder of the data contracts platform Gable.ai. Anderson says curation and management of data is crucial when companies use their data in AI applications. If you’re interested to find out how data contracts can make this happen then you’ll enjoy this interesting interview.
Sam had an engaging conversation with David Palmer, CPO of PairPoint by Vodafone and Web3 Asia Alliance board member about how Blockchain, Web3, and AI will redefine our perception of time and human interaction.
AI Events
The growing list of AI-themed industry events is evidence of the widespread disruptive effect AI is having.
This week saw the 8th Middle East Banking AI & Analytics Summit 2024 take place in Saudi Arabia.
The second annual DataConnect Conference West (#DCCWest) was held in Portland, Oregon where the vital contributions by women in AI and machine learning were in focus.
J On The Beach 2024 kicked off in Málaga, Spain this week. If you think that Big Data, AI, and a beach party in Spain are a good combo then there’s still time to sign up for tickets.
In other news…
Here are some other clickworthy AI stories we enjoyed this week:
- AI Is helping automate one of the world’s most gruesome jobs.
- Microsoft taps Sanctuary AI for general-purpose robot research.
- Andrej Karpathy suggests AI chatbots as interstellar messengers to alien civilizations.
- Sam Altman says GPT-4 is the dumbest model any of you will ever have to use.
- OpenAI working on new AI image detection and audio watermarking tools.
- Some say AI will make war more humane. Israel’s war in Gaza shows the opposite.
This is why AI will defeat all humans.#meme #memes #ai #humor #funny #lol pic.twitter.com/e3w1hzjJnQ
— Dan Ramos (@DanRamo07555230) May 3, 2024
And that’s a wrap.
Have you tried fully immersive VR? How do you know you aren’t experiencing it right now? Just kidding.
This reality is probably not a simulation and we should probably make more of an effort to interact with the other characters, I mean people.
Do you use AI transcription tools? Let us know if Whisper added interesting hallucinated words to your speech. The examples we shared in our article were pretty dark.
If you’ve been enjoying our interviews we’d love to know. Send us the names of the people you’d like us to interview next, along with links to any juicy AI stories we may have missed.
