• Google Health inks its first licensing deal for its AI…
  • The next-generation AI for text-to-image…
  • A video mapping start-up that reminds me of Waze…

Dear Reader,

It ended in applause. Applause?!?

One of the most-anticipated conferences in a long time happened yesterday.

It was The New York Times’ DealBook Summit, touting big names like U.S. Treasury Secretary Janet Yellen, Ukraine President Volodymyr Zelenskyy, Meta CEO Mark Zuckerberg, 48th U.S. Vice President Mike Pence, Founder of Netflix Reed Hastings, Amazon CEO Andy Jassy, and so many more.

But none of them were the main attraction.

Teleconferencing in from his cushy multimillion-dollar pad in the Bahamas was none other than Sam Bankman-Fried (SBF) of bankrupt cryptocurrency exchange FTX.

SBF at the DealBook Summit

Source: The New York Times

Given the events that have unfolded over the last couple of weeks surrounding FTX and SBF, there was much speculation as to whether SBF would show up – even if remotely. 

The level of corruption and fraud that’s led to bankruptcy and the loss of billions of dollars’ worth of customer funds gets even more unbelievable with every passing day. Yet the media coverage of SBF and what he’s done has been befuddling, if not outright bizarre.

The New York Times itself put out a head-scratcher of an article titled: “How Sam Bankman-Fried’s Crypto Empire Collapsed,” which didn’t mention once – or even suggest – that there was fraudulent, criminal, or shady dealings happening at FTX by SBF himself.

Shortly after that, The Washington Post outdid The New York Times with: “FTX collapse dooms founder’s effort to prevent another pandemic.”

That title was subsequently changed to: “Before FTX collapse, founder poured millions into pandemic prevention.”

The Washington Post article lamented that the money train had clearly stopped, and that FTX’s collapse would no longer allow SBF to pursue his magnanimous ambitions. It basically celebrated SBF and shared its biggest disappointment that there would be no more charitable-giving from SBF.

Is this journalism today? Millions have been defrauded, and that’s what The Post has to say? Pathetic.

And even The Wall Street Journal – a publication we assume would be all over one of the largest financial crimes in history – led with this one: “Sam Bankman-Fried Said He Would Give Away Billions. Broken Promises Are All That’s Left.”

That was accompanied by an equally ridiculous subtitle of: “FTX founder pledged to donate billions. His firm’s swift collapse wiped out his wealth and ambitious philanthropic endeavors.”

The Wall Street Journal mourned the absence of future philanthropic-giving from SBF and leaned into SBF’s belief in something called “effective altruism” – a concept of earning a whole lot of money so that you can give it back to meaningful causes.

It sounds great, of course, and it’s actually something that motivates me personally. But for SBF, it was all just a sham.

Below is an actual text from SBF from November 15 that was made public by a reporter at Vox who SBF thought was a friend. SBF’s comments are on the left, in gray:

SBF’s Private Conversation

Source: vox.com

There was a whole lot more, but it’s easy to see from the above that this was all just a sham. It was a lie. It was all window-dressing for what we now know is one of the biggest financial frauds of all time.

The interview conducted yesterday between SBF and Andrew Ross Sorkin of The New York Times was disappointing, much as I suspected it would be. After all, two weeks into this debacle, we can be certain that SBF has been well-coached by legal counsel.

The whole interview was so blatantly dishonest that it’s not worth recounting in any detail. And it can be summed with a few direct quotes from SBF himself:

  • “I didn’t ever try to commit fraud on anyone.”

  • “I didn’t knowingly commingle funds.”

  • “I’ve limited access to date.”

  • “I wasn’t running Alameda; I didn’t know exactly what was going on. I didn’t know the size of their position.”

When asked when the commingling of assets began: “So they didn’t, you know, lots of traders had open margin positions on FTX.”

When asked about criminal liability: “I mean, like, I’ve had a bad month.”

  • “I mean, like, look, I screwed up. We messed up big.”

  • “It was as truthful as I… you know, I’m knowledgeable to be.”

  • “I don’t know of times when I lied…”

SBF’s defense is so obvious, simple, and laughable at the same time. He claims to have told the truth, meant well, not taken customer funds, not known about what Alameda Research was doing (a firm that he owned 90% of), and just screwed up.

Yet it’s already well-known that he “borrowed” – actually stole – his customers’ funds, loaned them to Alameda Research for risky trading, and then Alameda took part of those funds ($1 billion) and loaned those directly to SBF.

We also know that SBF was the Democratic Party’s second-largest donor of almost $40 million this year, only behind George Soros himself. And it turns out that he also gave smaller amounts of money to the Republican Party – hidden in dark pools – so that he wouldn’t risk being “cancelled.”

And SBF was also public about his plans to commit somewhere between $100 million and $1 billion toward the 2024 presidential election.

There is so much more “dirt.” There’s an incredible web of commingling, appropriation of customer funds, outright fraud, potential money laundering, and misrepresentation – all while taking in billions from sophisticated and unsophisticated investors and living large in a $39.5 million Nassau, Bahamas penthouse. Wow!

And the audience applauded him at the conference. Unbelievable.

My bet is that SBF doesn’t serve a single day in prison. He bought too many favors with other peoples’ money, and that will keep him out of jail despite having committed one of the largest financial crimes in history.

Google continues its push into healthcare…

Google Health just made a first-of-its-kind announcement. It just inked a licensing deal with medical technology (MedTech) company iCAD for its artificial intelligence (AI) software.

What we’re seeing here is Google’s ambitions in the healthcare industry coming to fruition.

Regular readers may remember how Google has been striking deals with large healthcare systems over the last two years or so. These deals gave Google access to millions of patient medical records.

Well, Google has been using that data to train its AI software. And it’s starting with an AI that specializes in identifying breast cancer in mammograms.

We first talked about this back in 2020. At the time, Google’s AI demonstrated that it could reduce false negatives by 9.4%. And it reduced false positives by almost 6%.

So Google’s AI outperformed board-certified radiologists at the time. And it’s only gotten better. The AI has been training on larger data sets of mammograms, and that’s led to improvements in the AI’s accuracy.

And this is why iCAD just stepped up to license this technology. The company is going to turn around and use Google’s AI at more than 7,500 mammography sites in operation today.

That’s huge. This is Google’s plan playing out perfectly.

And get this: Google’s AI will continue to get better with every mammogram it analyzes through this deal with iCAD. Having access to so many new sites and imaging will only help the AI improve.

That’s certainly good for patients. It will improve patient outcomes by catching cancer that might be missed and can help to avoid giving some patients a false “scare.”

Of course, this is great for Google as well. Licensing deals like this will drive more revenue for Google’s Cloud division. Running a powerful AI requires computing power, and that means additional revenues for Google Cloud.

Right now, Google Cloud is a relatively small part of Google’s overall business. It only makes up about 7.5% of Google’s revenues. The other 92.5% is almost exclusively advertising revenue.

This is an area of focus for Google. There’s a lot of regulatory scrutiny over its practices around data collection for the purposes of generating advertising revenues. That’s not the case with Google’s cloud-based business.

Plus, Google will still collect all the patient data from the mammograms its AI assesses. This will help it build even more robust profiles of individuals. And that, in turn, will drive more advertising revenues.

So Google’s foray into healthcare is working out exactly as planned. And Google’s tech will certainly lead to better patient outcomes. 

It’s a bit of a double-edged sword, but that’s the point of Google’s strategy. Google uses the “good” that it’s doing to get away with its data surveillance and capture practices for advertising purposes.

From first to second-generation AI in less than a year…

Generative text-to-image AI has been the single hottest topic in artificial intelligence this year.

Just yesterday, we talked about how this technology is powering MyHeritage’s AI Time Machine tool. And we’ve also discussed all the major players in recent months.

Stable DiffusionGoogleMetaMidjourneyOpenAI’s DALL-E – there are a whole host of great generative AI tools out there. And each have their own strengths and weaknesses.

If we remember, it’s Stable Diffusion that’s powering the AI Time Machine we highlighted yesterday. And the company behind this tech – Stability AI – just released Stable Diffusion 2.0. This is a major upgrade.

For starters, Stable Diffusion’s image resolution was upscaled by a factor of four. That means the images the AI generates are now much higher-quality.

And it made some big improvements with regard to depth in the images. The AI now produces images that are incredibly photorealistic. The lighting, shadowing, and depth of images has really improved.

And Stability AI also made great progress with something called “inpainting.” Inpainting is an AI’s ability to basically fill in missing pieces, or pixels, when an image is changed or removed.

For example, if a creator doesn’t like a specific part of an image, they can cut it out. Then they can have the AI backfill the empty space with pixels that are consistent and contiguous with the image. This gives creators total control over the final product.

Here are a few examples of just how remarkable Stable Diffusion 2.0 is:

Stable Diffusion Images

Source: Playground AI

This snapshot shows off the AI’s ability to work with just about any style described in a text prompt.

For those interested, you can take a look around Playground AI to get a feel for what Stable Diffusion 2.0 is capable of. Some of the images are incredibly photorealistic. Others bring the most bizarre images to life. 

The most interesting thing to explore is the text prompt that was used to create the image:

AI Text Prompt

Source: Playground AI

We’ve explored the task of “prompt engineering” earlier this year in The Bleeding Edge

For the image above, it starts with: “Tiger floating in nebula.” Then there are some descriptions of the tiger’s feature. And there’s also reference to the hyperrealism style, as well as a particular artist.

As we’ve discussed before, writing prompts for an AI is a skillset itself. In fact, I wouldn’t be surprised to see an entirely new job title form here: Prompt engineer.

And the fact is, I’ve never seen software-based technology get adopted as fast as this is. Stable Diffusion 2.0 has been out for just a few days, and it’s already one of the most actively used software on the GitHub repository.

The utilization of this kind of tech will be a massive trend in 2023. It’s going to transform the industry for creators, graphic designers, and also for content producers.

A hot early stage company to keep an eye on…

We’ll wrap up today with an early stage company doing some interesting work. The company is called Nexar, and it reminds me a lot of Waze.

I suspect most readers are familiar with Waze. It’s the mapping and navigation app that’s been around for over a decade now. I’m sure that most of us have used it at one time or another.

Waze launched with one of the most brilliant business models I’ve ever seen. It started with the free mapping application. That itself wasn’t unique.

But unlike its competition, Waze constantly monitored all of its users’ driving conditions in real-time. This gave it instant insight into traffic patterns, which in turn improved its navigation.

For example, if a bunch of drivers using Waze slowed down or came to a halt on one part of the freeway, Waze would update its map to reflect this. And it would route users around the problem area.

Then Waze even built in functionality for users to report accidents, construction, and traffic jams themselves. This made the navigation system even more robust.

And because Waze effectively crowdsourced all of this information, the app was relatively cheap to produce and maintain.

Of course, what users didn’t know in the early days was that Waze also tracked all of their data – even when they weren’t using the app.

Waze documented everybody’s travel patterns and preferences. And it knew exactly where everybody was at all times.

This made it a goldmine for advertisers. And sure enough, Google came in and acquired Waze for nearly $1.2 billion in 2013. It was the perfect fit for Google’s business.

Well, Nexar is following the same model. It’s developed a network of more than 700,000 vehicles that have outward-facing cameras on them. These cameras are collecting video footage of the roads and driving conditions in real-time.

All of this video footage is crowdsourced. And Nexar is applying its own AI to all this data to essentially optimize logistics and delivery networks.

The idea here is that logistics companies can use Nexar’s network to route and re-route their delivery fleets to maximize efficiency.

Nexar has the ability to not only collect the data, but literally see what’s happening on surface roads in real-time. It can know what the cause of a traffic jam is or what kind of construction might be taking place on surface roads.

What’s more, this video footage gives Nexar insight into the driving habits of every driver in its network. And guess what?

The venture capital (VC) arms of both State Farm and Nationwide are early stage investors in Nexar.

Clearly, the insurance industry thinks that Nexar’s tech can help it optimize its actuarial practices. If they can get data that defines a driver’s specific habits, they can price their auto insurance policies much more efficiently.

While Nexar is enterprise-focused right now, it’s easy to see how there can be even wider consumer applications of its software and data. Will Nexar become the next Waze? And who will acquire it? Or will it resist and remain independent?


Jeff Brown
Editor, The Bleeding Edge