This week, we’re talking about AI that attends meetings for you, Apple is spending millions of dollars per day on AI, X, formerly Twitter, now can use your posts to train its models, the Pentagon is investing heavily into an AI drone network, robots have been trained to outsmart humans at tag, an AI surgeon is saving lives, the largest open-source AI model is released, a tiny country is benefitting tremendously from the ai gold rush, and so much more!
Google Meet AI
Are you tired of paying attention in meetings or needing to show up at all? New AI from Google might make that a thing of the past. This week at Google’s Cloud Next conference, Google launched a suite of new functionality for Google Meet, their video conferencing software. First, with the click of a button, Google will take notes and create action items automatically. And, if you’re late, it can also show you a mid-meeting summary to catch you up. You’ll be able to talk with Google’s bot privately during the meeting, and when the meeting is over, it’ll send you all the notes and include video clips of the most critical moments. And this all sounds good, but why even show up at a certain point? Google Meet will also have an “attend for me” button that auto-generates text about what you might want to talk about and provides it to the attendees. And not to be outdone by Google, Zoom is essentially launching the same features with a Zoom AI Companion product with similar meeting summary, mid-meeting catch up, and other features. So, who thinks the future will be just a bunch of AI bots talking to each other?
Robots Learn To Play Tag
Next, researchers at UC Berkeley have trained robots to outsmart humans at tag. This is not a simple following algorithm. The robot actually cuts people off and actively searches for them. This is all based on a new way to teach AI strategic decision-making for dynamic tasks such as playing tag. According to The Decoder, “Learning such behaviors in the real world is extremely difficult for a robot because it has limited knowledge of its environment and other agents through its sensors, the goals of other agents are unclear, and movement in the physical world is fundamentally more difficult than in simulations. Therefore, direct learning of such behaviors, for example through reinforcement learning, has failed due to these requirements.” This novel new approach, privileged learning, aims to help AI understand the real world more effectively during decision-making. So now robots will be able to hunt us down much more effectively, one step closer to the Terminator.
Apple AI Spending
Now, let’s talk about the two biggest tech companies in the world. First, Apple is rumored to be spending millions of dollars per day training models. According to Mac Rumors, Apple’s AI Chief John Giannandrea is said to be skeptical of AI chatbots, but even with that skepticism, he formed a team to work on conversational AI four years ago. We previously talked about Apple’s internal AI chatbot codenamed Ajax. The team behind it is only 16 people, but they spend millions of dollars per day training large language models. But text-based chatbots are not what Apple believes is the future. Instead, they continue to bet big on voice-based AI, aka Siri. A voice interface with AI always made a lot of sense to me and, in my opinion, will be the interaction method that takes AI to the next level. And it’s not like we don’t have the technology to do that right now. But in true Apple fashion, they want to make the experience unique to Apple. Apple wants to give Siri the ability to execute multi-step commands. The pieces of this technology are here already, between open interpreter which I just published a video about, large language models, translation AI, voice-cloning, and transcription. It’s possible to put this together right now.
Microsoft AI Backpack
And for the second biggest company in the world by market cap, Microsoft has filed a patent for AI-assisted wearables, including a backpack. According to WindowsCentral.com, “The patent illustrates how a backpack with sensors could relay information from a user’s surroundings to an artificial intelligence engine to provide assistance to a person.” One example given in the patent application is a skier deciding where to go when the backpack scans the area and lets the skier know a particular area is out of bounds on the mountain. Microsoft is betting big on AI with their investment in OpenAI and AI integration into seemingly every software product they offer. However, one category they’ve historically not done well with is hardware, except for Xbox. They might view this as an opportunity to be a strong contender in the hardware markets, but there’s an equally strong chance the product described in this patent never sees the light of day.
Falcon 180b Released
Speaking of large language models, there’s a new king of the open-source world. Falcon, a unique foundational model similar to ChatGPT and LLama, has released a new 180 billion parameter version trained on a massive 3.5 trillion tokens. This is more than twice the size of the next largest open-source model, LLama 2 70b. Falcon’s new version immediately jumped to the top of the open-source LLM leaderboards. But as this is just a foundational model, further fine-tuning will be required to turn it into a powerful conversational or instruction-based model. I love seeing open-source foundational models other than LLaMA. Competition is great for consumers. The Technology Innovation Institute created this model out of the UAE, and like its predecessor, it continues to be open-source, free, and commercially viable.
AI Hearing Aids
Next, in health and medicine, AI has been making tremendous strides in improving people’s well-being. First, thanks to AI, hearing aids may soon offer translation built-in. A Minneapolis-based company uses AI to give hearing aids advanced features like real-time language translation, fall detection, and alerts. Hearing loss affects 40 million people in the US, and they could miss more sounds than they realize. Now, with this nearing hearing aid tech powered by AI, they could hear better and have super-powered hearing beyond what ordinary people have. And hearing loss affects not only older people but 1 in 7 teenagers now due to things like earpods. This seems like a combination of a hearing aid and something futuristic like the Babelfish.
AI Brain Implant
Also, in AI health news, Ph.D. students at UC San Francisco have developed hardware that allows people with communication challenges, such as stroke survivors, to communicate more naturally through a digital avatar by turning brain signals into text, speech, and facial expressions. Ann, a stroke survivor, had a device screwed into her skull which reads her brain signals, allowing such communication to occur, a first of its kind. I love hearing about all the AI health advancements, it gives me hope for the future.
AI Robot Surgeon
Our last AI health story is about an AI robot surgeon who was able to operate on a historically inoperable tumor, saving the patient’s life. This robot, called Davinci, can operate in ways humans cannot, for example, reaching areas that human hands can’t. The robot itself isn’t new, but now it has the power of AI behind it. Prior to the introduction of Davinci, the surgery would have been much more severe, requiring a more aggressive incision and a significantly more extended recovery period.
X Is Using Your Data
It seems Elon Musk doesn’t want anyone else to be able to use AI but him and his companies. Honestly, Musk might be the most contradictory person who ever lived. In a new update to their terms and conditions, X, formerly Twitter, allows the company to use your posts and information to train their AI models. Since taking over Twitter, Musk has gone to incredible lengths in building a moat around Twitter’s data. He started by cutting off API access and then formed a separate entity xAI, which he says is a separate company but will work closely with X. Now it’s becoming more apparent how that will manifest. By cutting off access to the data, he’ll likely provide his own AI company, xAI, with direct and unfiltered access. And now they will legally be able to do so since they changed their terms and conditions to allow it. I made an entire video about how AI is changing the way the internet works by forcing companies with large amounts of user-generated content to shut down 3rd party access to that data. When a company's data is so incredibly valuable, the natural competitive inclination is to hoard it all for itself.
US Chip Restrictions
AI and geopolitics are coming crashing together. According to Reuters, “The U.S. expanded the restriction of exports of sophisticated Nvidia and Advanced Micro Devices artificial-intelligence chips beyond China to other regions, including some countries in the Middle East.” Restrictions like these are usually put in place due to security concerns, which accompany the growing problem the US has about China’s technological capabilities. In parallel, the US has invested heavily into moving chip manufacturing to US-soil, decreasing our reliance on other countries in what is becoming a critical commodity.
Russian Supercomputer
And in the same realm, a new Russian AI supercomputer has come online with 400 PetaFLOPs of power. According to Tom’s Hardware, “Lomonosov Moscow State University (MSU) has launched its new supercomputer named MSU-270…will be used for various artificial intelligence (AI) and high-performance computing (HPC) applications and for training large AI models” Governments around the world are building supercomputers to win the AI race and gain AI supremacy. But, there’s also a darker side of how this can all end. How do you think it will end?
AI Drone Fleet
And the US is investing just as heavily in AI for defense. This week, the Wall Street Journal states, “The Pentagon intends to field a vast network of AI-powered technology, drones and autonomous systems within the next two years to counter threats from China and other adversaries.” The US government spends hundreds of millions of dollars on air, land, and sea-based AI systems. Kathleen Hicks, the deputy secretary of defense, said, “Imagine distributed pods of self-propelled [autonomous] systems afloat, powered by the sun and other virtually limitless resources, packed with sensors aplenty, enough to give us new, reliable sources of information in near-real-time” So yea, autonomous AI military robots are going to be all around us…cool cool.
Crypto AI Scams
Our next story is one that we’ve touched on in previous videos. Each fantastic new AI innovation almost immediately gets translated into technology a scammer can use to defraud people. According to the Cointelegraph: “Quantstamp's Richard Ma explained that while social engineering attacks have been around for some time, AI is helping hackers become “a lot more convincing” and increase the success rate of their attacks.” The quality of these attacks is increasing, and the scale is also due to AI. Persuasive social engineering attacks are made much more possible for anyone to commit since not much thinking and planning has to be done anymore. As previously reported, X is experiencing an increase in crypto-related scams, and I envision this will only worsen as AI improves. Just remember these two things: if it seems too good to be true, it probably is, and trust but verify.
AI Video Of The Week
Congrats to this week’s winner, Kytr.a!
Tiny Island AI Goldrush
Our last story of the day is about a small island in the Caribbean called Anguilla. This tropical island might have had an enormous cash cow fall in its lap because of the AI gold rush. Anguilla, a tropical British territory, is known as an island paradise. However, since the 1990’s, it has also been responsible for assigning domain names ending in .ai. You can probably guess where this is going. Now, it seems every AI startup is rushing to get domains ending in .ai, and for good reason. Most .com domains that are not complete misspellings or super long are long gone, and .ai is an excellent domain for ai companies. Stability.ai, X.ai, and Character.ai are just a few recent AI companies that have purchased domains from Anguilla. And the most prominent tech giants also have .ai domains, including Google.ai, Facebook.ai, and Microsoft.ai. According to Bloomberg, “The total number of registrations of sites ending with these two letters has effectively doubled in the past year to 287,432, and Anguilla will bring in as much as $30 million in domain registration fees for 2023.” It seems ChatGPT has changed this tiny island, according to Vince Cate, who manages the .ai domain names for Anguilla, who said, “Since November 30, things are very different here.” And I’m one of the people buying a .ai domain name. I’m launching an AI consultancy soon and purchased a domain ending in .ai for my company’s website. But, more on that soon enough.
