Issue 9 - The Emerging Ecosystem of Generative AI 📈

The Emerging Ecosystem of Generative AI

Large language models, or LLMs are having a massive impact on our society, in particular the way we communicate and process information. The most notorious or famous is OpenAI’s ChatGPT. However, this past week saw news about competing models that demonstrates how active and dynamic this emerging industry is.

In many respects, Generative AI is the evolution of search engines, making it easier for people to make sense of the massive volumes of knowledge and media we now have access to. A big difference between existing search engines and these emerging services is that the latter are largely behind paywalls, while the former were free and subsidized by advertising.

LLMs are basically statistical engines that can predict text. To be clear, they do not understand or comprehend the text, rather they can predict what text matches a particular query or question. The same way a search engine can predict the web result you’re after, so too can LLMs predict the text result you are after. Where a search engine can give you thousands of web results, an LLM can give you thousands of words as a result of your query.

These predictions are determined by whatever media has been analyzed to build the model, and increasingly large language models are using as much text as they can gather. An effective web search engine has crawled millions if not billions of web pages, and an effective LLM has similarly processed billions of lines of text. Similar logic applies to images, video, or audio.

While OpenAI is an industry leader, there is some criticism and ethical concerns around what text they use, and how they obtained it. Authors and other copyright holders are understandably concerned that their work was used to train these models without their consent.

Enter Twitter, the company now known as X, and their LLM called Grok:

While Grok seems to embody the rather toxic culture that Twitter is now embracing, it also has a clear advantage over other LLMs: access to Twitter’s data. This instantly positions Grok as more current and diverse than other models, given the range of content and news posted on Twitter. Mind you, accuracy may be an obvious issue with Grok, as disinformation on Twitter flourishes under the current (mis)management.

Where Twitter has an advantage due to their data, other companies are leveraging the world of open source to gain both speed and scale. Take for example the company 01.ai

Not only is 01.ai using open source software, but the company offers their chat bot in both English and Mandarin, providing a kind of AI bridge between the two languages. Businesses who operate in both the US and China might see advantages in using a model that combines the words and cultures of both. However this also ensures that the company falls squarely into a larger geopolitical conflict.

It is also worth noting that Facebook (o/k/a Meta) is also using open source to keep up with the Generative AI race. Unlike Twitter, it is not clear how much of their user data is going into it, although it is safe to assume that this is part of their competitive advantage.

WhatsApp and Facebook Messenger are where Meta sees a lot of their chat bots residing, and the company already allows certain partners to build their own custom version of these chat bots.

Not to be outdone, OpenAI this week announced that any paying customer can create their own GPT, and make their version available in an app store to compete with each other. Think of it as less of a custom chat bot a la Facebook, and more a custom search engine. Only this custom search engine can then be made available to others, competing for the most effective configuration.

For example, if you feel ChatGPT is too biased or not biased enough, you can configure your own version, and if it is popular, get paid by OpenAI.

As with any hype cycle, there are ups and downs, and with Generative AI, the hype is definitely climbing to new heights. Yet if we move past the hype, we can see real substance when it comes to how we find and organize information. Hence given the utility of search, if this represents the future of search, there really are trillions of dollars at stake. Especially when you factor in the cloud computing business that drives all of this.

How all of this plays out is not yet clear, but the speed is significant. We’ll continue to monitor developments and update our fellow punks.

Notable Punks: Nzou (Zoe), the Buffalant 

Loneliness sucks. Everybody knows that. It sucks to an extent that the term is oftentimes followed by the word “pandemic” because it apparently even promotes an early death.

So far, not even Ozempic claims to have found a cure for it. Fortunately for us humans, and unfortunately for our animal friends, loneliness is a phenomenon that is inclusive of all species in today’s society, in particular for those who are released from captivity into what we optimistically refer to as freedom.

Not all stories end as tragically as Kiko’s (who seemingly couldn’t shed the confusion of not being with humans anymore and eventually died of pneumonia). Let’s take a look at Zoe, the elephant turned buffalant, and learn from a masterclass in adaptability, flexibility, persistence, grit and - yes, it doesn’t sound pretty but needs to be said - pretense and fake-it-till-you-make-it.

Zoe, now 50 years old, was orphaned at a reserve in Zimbabwe and then paired with the largest non-elephant species in the park, the buffaloes, who she immediately took a liking to. In fact, she liked the buffaloes so much that she swiftly took over the herd and established a firm, merciless matriarchy amongst the buffaloes (so far, only 14 males who tried to challenge her discovered that doing so was a fatal mistake, but that number will certainly go up in the coming years). After 7 failed attempts to re-integrate Zoe with elephants, the park rangers finally gave up and let Zoe now be whoever or whatever she wants to be or thinks she is. At the point of this writing, it is unclear whether Zoe self-identifies as a buffalo or if the buffaloes now self-identify as elephants and, frankly, it neither should nor does matter. What does matter is that Zoe successfully escaped loneliness, and that both her and her herd have found a way to (more or less) peacefully live in harmony with one another.

Talk about taking the road less traveled! If anyone ever was deserving of the title “trail-blazer”, it most definitely and absolutely is Zoe. What a punk. We can all learn a lot  from so much flexibility and open mindedness, as we can learn from elephants in general.

Be sure to find your buffaloes!