Last updated on November 9th, 2023 at 10:55 am
We are watching the beginning of the AI Ecosystem Wars.
There have been skirmishes, but competition launched in earnest this week when Elon Musk announced Grok AI the day before Sam Altman and Satya Nadella talked onstage about how ChatGPT was overhauling Azure.
It’s pretty wild. Five companies whose tech and ecosystems make them the only companies with a serious shot at leading the most significant change humanity has ever experienced.
The Big 5 AI Ecostems
1. OpenAI’s Ecosystem: OpenAI has built an ecosystem centred around groundbreaking AI research and applications, with flagship products like ChatGPT and DALL-E that are at the forefront of conversational AI and creative algorithms. Its ecosystem is bolstered by a strategic partnership with Microsoft, providing a robust cloud infrastructure and research collaboration that drives innovation in AI.
2. Google’s AI ecosystem is expansive and deeply integrated into its suite of products and services. Its new Gemini model is expected any time now. It was expected to be the first multimodal model until ChatGPT beat it in November. It includes:
– DeepMind Technologies: A leader in AI research, working on projects like protein folding with AlphaFold and AI for health.
– Google Cloud AI: Provides AI and machine learning services to businesses, supporting a range of applications from data analytics to AI-powered efficiency.
– Google Research: Focuses on advancing AI and machine learning across various domains, contributing to both academic and practical applications.
– Consumer Applications: Implements AI in products like Google Assistant, Google Photos, and YouTube for personalized user experiences.
– Android: Employs AI for system optimizations and app functionalities on mobile devices.
– Hardware: Develops AI-optimized processors like TPUs to enhance performance in large-scale computing tasks.
Google’s ecosystem also includes Midjourney, one of the most powerful and popular AI image-generation engines. Its partnerships span academia, industry, and technology sectors, continuously fueling its AI ecosystem with innovation and collaborative research.
3. Meta’s Ecosystem: Meta has crafted an AI ecosystem that powers social networking and virtual reality experiences, with AI-driven technologies enhancing user engagement and content discovery. Meta’s AI research division contributes to the broader AI community while driving internal product development.
4. Amazon’s Ecosystem: Amazon’s AI ecosystem includes ChatGPT’s biggest competitor, Anthropic, and Stability AI, an open-source generative AI model that encompasses a vast array of services, including cloud computing, consumer products, and retail operations, all powered by AI. AWS offers AI solutions including hosted LLMs to businesses, and Amazon’s consumer tech like Alexa and holdings like Whole Foods bring AI into homes, illustrating the company’s diverse application of AI technologies.
5. Apple’s Ecosystem: Apple’s AI ecosystem integrates across its hardware and software offerings, with AI features like Siri and advanced image processing being central to the user experience. Apple’s commitment to privacy is a defining trait of its AI, with a focus on on-device processing. Through collaboration with chip manufacturers and a developer-friendly platform, Apple ensures its ecosystem remains innovative and secure.
So who has pole position? Right now there’s no contest, OpenAI’s Microsoft relationship is defining the space. Google has the depth but has taken its foot off the gas. Musk has all the ingredients, nearly.
What about Tesla?
I have the halcyon vision of a gentle robot helping a 98-year-old lady into bed then washing the dishes and knitting a sweater all night long. You can see this is coming. It’s partly why Musk’s work at Tesla is so important; the kind of 3D awareness that will be necessary is very far away, but much less far than it once was. It will come, as will the physical/AI integration, something else Musk is sort of working on at Neuralink. I confess to some surprise he hasn’t moved into robotics yet l. When he does he’ll have vertically integrated the next wave of the future.
But I digress, seeing as quality of life for the vulnerable and underprivileged hasn’t exactly been a focus for him to date.
What he has done is build the first and only closed, proprietary AI network.
– Tesla: The world’s biggest set of mobile, proprietary supercomputers. A vast data set from its fleet of vehicles, feeds into its AI systems for autonomous driving.
– Dojo: A powerful supercomputer for AI processing.
– SpaceX: Utilizes AI for various functions, including satellite deployment and possibly operations. Generates data, with high processing power.
– Neuralink: potential advancements in AI interfacing with the human brain.
– X: an exclusive, high-value, unique dataset with high training potential likely to lead to much better results than an AI weaned on romance novels.
With Twitter, now known as ‘X’, Musk gains access to a wealth of social media data, which could potentially be used to train AI systems to understand human communication patterns, social dynamics, and more.
The potential behind this network is enormous, unique and terrifying, especially given Musk’s mercuriality and right-wing coziness. Despite advising caution, he’s not one to put the breaks on. Clearly. The thought of a world-changing technology under his sole control is deeply unnerving.
Other assets add to his ecosystem strength.
Team: a fleet of AI engineers and developers working on the frontier
Experience: so much experience managing AI projects should not be discounted
Focus: Because it is he and he alone, he is unfettered by boardroom debate or any except the most essential conversations. He does what he wants. When you have the infrastructure and complete control, you can focus very selectively and very effectively.
While Musk’s AI network is undeniably powerful, addressing AI’s ethical use is as crucial as its technical prowess. For instance, Tesla’s autopilot system raises questions about safety and accountability that are yet to be fully resolved. To compete or collaborate in this ecosystem, companies must not only harness data but also cultivate trust through transparent and responsible AI development. As AI evolves, a balance between innovation and regulation will be essential to ensure technologies like Neuralink and autonomous vehicles benefit society at large. Ultimately, the success of AI may not just be in speed or data but in fostering a harmonious coexistence with human values.
Meeting AI’s Challenges
What will define ecosystem success next? Accuracy. What OpenAI showed yesterday was extremely impressive. But there is deep, almost existential concern about accuracy. So far, Generative AI is utterly brilliant but kinda lazy, error-prone, and not too concerned about facts. How did that happen? We can only look in one place and that is the training data.
AI IS data. So if it is behaving poorly, it’s in the data somewhere, why?
Trained on Fiction
I have a crazy theory. If you brought an alien down to earth and had them do nothing but read fiction for a year, it would have a pretty skewed view of humanity. If you had an alien ingest data from which it can make associations and assumptions, that informed its behaviour … if it starts acting like an asshole, we have only ourselves to blame.
It’s artificial, but it’s us. Every iota of data in these platforms’ virtual brains has come from us. That’s why it’s racist. It’s not like AI just started downranking spontaneously. And not only is it from us, it’s stories about the very worst and best of us, from which is it possible to glean information about how to behave.
How else would an AI learn to be lazy? Because it is. Instead of verifying info, it makes it up. It’s got the attitude of a pot-smoking, shrooms-ingesting teenager in your basement. Remind you of anyone? How about … the antihero?
I’m not saying AI is creating a code of behaviour based on the fiction it ingests. But hallucinations are real, and possibly rogue, and possibly signs of incipient intelligence. But right now they are just a huge pain. “The thing about working with LLMs is the demos are amazing but the actual day-to-day outputs makes mistakes or vary unpredictably often enough that it’s tough to coerce them into doing the right thing consistently and repeatably.” someone in the industry told me today. So the search for the next model or the corrective tissue or the antidote content continues.
Is the LLM Well Poisoned?
And yet we seem to expect a higher standard of behaviour without teaching it why the data we gave it may have led in the wrong direction.
Someone asked me today if an LLM could be trained in antifacist techniques. Yes. An LLM can be trained to do anything with sufficient quantities of the right content. But if the accuracy issue isn’t addressed soon, real questions of value are going to start to crop up.
And other questions. Is this and the Hinton letter why Google is just absent in this ecosystem build-up right now? Is every ecosystem now centred around data? Can anyone catch up to OpenAI and Musk?
Go Fast, or Go Far
If I know anything it’s that this is a loooooooong game. There’s an old saying: if you want to go fast go alone. Musk has caught up fast and he will likely continue to move faster than anyone else.
The rest of the saying goes: But if you want to go far, go together. The solidity of the OpenAI / Microsoft relationship is heartening to see
After the sturm und drang inevitable from MuskAI, we’ll be craving comfort food.
The landscape is increasingly complex and important for people to visualize and monitor how all these companies and technologies intersect. This week, watch for the first draft of the first edition of the AI ecosystem, 2023. To purchase the full version and accompanying commentary, visit PatternPulse.ai
Written with support from ChatGPT 4.0. Illustration by Jen Evans and DALL-E
