Dear Reader,
A few days ago, Italy announced a ban on OpenAI’s ChatGPT because of a range of privacy violations.
It was just a matter of time.
Given the controversy surrounding such a powerful generative artificial intelligence (AI), it was inevitable that a government would take issue with the new software. There will be more that follow.
It’s well known that OpenAI’s ChatGPT violates the EU’s General Data Protection Regulations (GDPR). While OpenAI’s official statement is that “we believe we comply with GDPR,” OpenAI does collect data and conversations with ChatGPT in order to train the AI.
It also didn’t help that OpenAI had a data breach with ChatGPT last month that revealed some users’ chat histories with the AI as well as payment related information for about 1.2% of ChatGPT Plus subscribers.
And perhaps even more salient is the fact that OpenAI doesn’t have any age verification for its users. Children who experiment with or use ChatGPT can be exposed to completely inappropriate material for their age.
Italy has provided OpenAI 20 days to implement changes to address the country’s concerns. If it doesn’t do so, it could be levied a fine of up to 20 million Euro or 4% of worldwide annual revenue.
Is that a real deterrent for OpenAI? In reality, not at all. And it reveals OpenAI’s approach to rolling out this powerful new technology.
The thing is that Italy’s areas of concern are logical, and they are well known by both OpenAI and the industry. OpenAI knew of these issues but has not yet implemented changes to its software to enable age verification, or to stop using user interactions as a way to train the AI.
It could have held back the technology until those had been developed before releasing the software to the public, but it didn’t.
It’s approach to releasing the technology into the “wild” has been calculated. After all, the people behind OpenAI are some of the most sophisticated technologists and computer scientists in the world. They know exactly what they are doing. And they are optimizing for speed of development – plain and simple.
The team at OpenAI also knows the risks. And I’m certain that they’ve taken the approach of dealing with them when they come.
The reality is that the monetary risk is insignificant to OpenAI. After its recent $10 billion raise, a 20 million Euro fine is nothing. And it is so early in its monetization, even 4% of worldwide revenues wouldn’t be much at all.
This is a classic example of a Silicon Valley ethos to “move fast and break things.” When an entirely new industry is being created, it’s a land grab for domination.
And in the world of software, when nearly everyone has a computer and/or a smartphone, the battle for dominance can now be won in a matter of months, not years.
I’m not suggesting that this is the right approach. After all, age verification would be fairly simple to implement. And it would be easy to not use users’ interactions with ChatGPT for training purposes.
But from OpenAI’s perspective, that would slow things down. It would require far more effort to curate training data for a specific age group, and it would have to give up what is effectively free crowdsourcing of training its AI. And that means it would lose time in this furious race.
And time is money given the size of the economic opportunity to reinvent search and displace Google for the first time in more than 20 years..
Amidst the race to develop the best large language model (LLM) to compete against OpenAI’s ChatGPT and GPT-4, another race that is just as exciting has been running in parallel. This one is for text-to-image generators.
The technology is precisely what it sounds like. Users enter a written prompt, and the software creates an image based off that prompt.
And less than two weeks ago, Midjourney released Midjourney 5 in an effort to leap ahead of the other two large players, Stable Diffusion and OpenAI’s DALL-E 2.
The results are nothing short of amazing.
Midjourney 5 Creates Life-Like Images
Source: MidJourney
Shown above is an example of the incredibly realistic images that Midjourney 5 can produce.
The textures, lighting, depth, resolution, and shadowing are remarkable. It’s like seeing everything in high definition, as if every image is a professional photograph of the real world.
Part of the upgrade of Midjourney also includes an even wider range of stylistic flexibility and natural language understanding making the results even more fantastic.
While the default setting is photorealism, given the right prompts, Midjourney 5 can create an image in any style or artist’s likeness that we can imagine.
This latest release of Midjourney 5 was just an “alpha” version release. That means it is very much a work in progress despite the current capabilities of the artificial intelligence (AI). And it will only get better.
The demand to experiment with the new software was so overwhelming, Midjourney had to shut down its free trial a few days ago.
Perhaps not surprising, not only did Midjourney have an excessive demand problem, but the software was also being widely used to create deep fakes of well-known personalities.
An example of a highly realistic, but fake image that many have already seen was of Pope Francis wearing a trendy white puffer coat.
The AI-Generated “Pope Coat” From Midjourney 5
Source: Midjourney, Forbes
For a short period of time, this was widely thought to have been a real photograph until it was identified as being AI generated.
The giveaway, which most didn’t notice, is the Pope’s right hand in the image on the left. Image generators have struggled with hand generation, but it won’t be long before that is corrected.
The radical capabilities of these text-to-image generators have arrived far faster than most believed possible. And the improvements are happening in a matter of weeks, not years.
Our largest challenge with this kind of technology is trying to figure out what is real and what’s not. Ideally, AI generated images would be watermarked somehow so that we’ll be able to easily identify an image for what it is.
But this will never be able to stop bad actors from circulating deep fakes to push a narrative. This will make things that much harder for us to determine which information sources we can trust or not.
Microsoft recently released a new functionality in its Edge browser to its employees. And it comes as a bit of a surprise.
The latest Edge upgrade enables a cryptocurrency wallet accessible through the internet browser.
Microsoft is Experimenting with a Crypto Wallet in its Web Browser
Source: Twitter (@thebookisclosed)
It is a non-custodial cryptocurrency wallet which means the user has control over the wallet, not Microsoft. It is also capable of supporting payments, as well as holding non-fungible tokens (NFTs).
Implementing a crypto wallet into a browser is nothing new, but it has typically been done by using plugins. A perfect example is popular non-custodial wallet MetaMask which has plugins for Chrome, Firefox, Brave, Edge, and Opera.
What is unusual is for a company like Microsoft to be developing the cryptocurrency functionality itself. By doing so, it may put the company into regulatory crosshairs given the animosity the SEC and current administration has towards digital assets of any kind – other than a digital dollar controlled by the Federal Reserve, of course.
Perhaps Microsoft isn’t too concerned as Google has more than 10X the browser market compared to Microsoft in terms of daily active users. But as we’ve seen, that’s been changing given the success of its Bing search engine with OpenAI’s GPT-4.
I can’t help but wonder if this new crypto wallet functionality is a trojan horse to what will ultimately be digital dollar payments enabled within a browser.
There is great utility in embedding a payments mechanism into a browser, especially in a world where micropayments in exchange for content become commonplace.
This is something that becomes possible with digital money and traditional credit card payment rails are no longer required.
The largest surprise and embarrassment amidst all of these incredible developments in artificial intelligence over the last four months has come at the expense of Alphabet (Google).
After all, Alphabet has invested tens of billions of dollars into developing artificial intelligence and has long been positioning itself as an “AI company.”
And yet, OpenAI, Microsoft, Midjourney, Stable Diffusion, and others have been running circles around Alphabet, and spent a fraction as much in research and development.
Even more embarrassing is that a “faction” within Alphabet – DeepMind in the U.K. – has been delivering the world one AI breakthrough after the next. But its breakthroughs haven’t been in areas that would benefit Alphabet’s core business, search and advertising.
Which is why, late last month, Alphabet rolled out its large language model (LLM) called Bard – its competitive answer to Open AI’s GPT-4 which is used by Microsoft’s Bing – with very little fanfare.
The truth is that Alphabet is way behind…
Good luck gaining access as well. While it is technically available to users in the U.K. and U.S., it is available by waitlist and access must be granted to individual users.
Google is going to great lengths to stress that the rollout of Bard will be slow. It seems that the focus is to limit usage in order to buy some additional time to improve Bard until the company feels that it is ready for a commercial rollout.
After all, about the only thing worse than getting beat to market like Google has already, would be for Google to launch a product that was markedly worse than what is already out there.
This is a critical matter for Alphabet, and it could have massive consequences for the company’s stock.
If we remember, when Alphabet flubbed its first public demo of Bard, it cost the company $100 billion in lost market cap when the stock fell 9% on the news.
If its commercial release is also unimpressive, it could result in even larger losses for the company. This may feel like an overreaction, but in reality, it’s not at all.
Natural language processing, large language models, and generative AI are the key to the next generation of search.
And if we take one step further, it’s not hard to image a user interface that no longer requires an internet browser at all. These generative AIs, when fed with real time information, will be far more useful and time -saving than the way that we access the internet today.
They have the potential to replace the need for Google entirely. And if that were to happen, more than 80% of Google’s revenues, which are driven entirely by search related advertising, would be at risk.
Regards,
Jeff Brown
Editor, The Bleeding Edge
The Bleeding Edge is the only free newsletter that delivers daily insights and information from the high-tech world as well as topics and trends relevant to investments.
The Bleeding Edge is the only free newsletter that delivers daily insights and information from the high-tech world as well as topics and trends relevant to investments.
