AWS and Hugging Face bring GenAI to everyone
Plus the latest on the AI wars, censoring ChatGPT in China, and why AI-generated images are *not* protected by copyright
Happy Friday, all!
Another week of madness, and although it’s only February, it’s a great time to reflect on five of the ten predictions I made for Generative AI in 2023. They are taking shape even quicker than I had imagined.
🚀 First, on mass adoption, there is no doubt that this is GenAI’s year. GenAI products, tools and services are flooding the market. Every company I speak to is considering its GenAI strategy. There is a growing realisation that we are witnessing a step change in the economy and society.
📣 Second, on controversial public debates, it’s only a matter of time before the pendulum swings back from the enormous utility in the ‘business layer’ of GenAI to the unexpected and straight-up bad sh*** people are using these new tools for. The trickle of negative headlines we see today will become a torrent. It’s going to get intensely political.
⚔️ Third, on the AI Wars — I was among the first to predict how the battle between the tech giants trying to ‘win’ in this space would unfold. And unfolding it is. We’ll come back to this story soon.
🔥 Fourth, I predicted the David v Goliath dynamic — that the tech titans would try to compete or collaborate with the rising stars of the GenAI space. That’s playing out to the script. Microsoft went all in on Open AI; Google followed suit with Anthropic — and this week, a third tech giant made both a play and a collaboration. I think it’s hugely important… let’s get into it.
AI for all: Hugging Face and AWS
🤗 On Tuesday, Amazon’s cloud platform — AWS — the most broadly adopted cloud platform in the world, announced its collaboration with Hugging Face. If you haven’t heard of Hugging Face, make a mental note now. Set up six years ago, it’s evolved into an open-source platform and community, where users can access OS code and technologies to build, train and deploy machine learning models.
💵 Hugging Face is on a mission to make the best generative models — which require a lot of data, computing power and basically — widely accessible. By leveraging AWS as its preferred cloud provider, Hugging Face will allow developers to access AWS’s state-of-the-art tools to train the best generative models without needing bucketloads of moolah. A kind of ‘AI for the people.’
Why is this important?
👉 One, because while headlines are obsessed with Google and Microsoft — Amazon (or AWS) is gegenpressing. They’ve been collaborating with Hugging Face for some time — this solidifies that base. Other tech giants will be doing ‘stuff’ too. (Hello, Apple?🍎)
👉 Two, because this partnership aims to democratise Generative AI. (To be clear, AWS isn’t entering into this partnership out of the ‘goodness of its heart— it is also a means for AWS to secure the dominance of its cloud platform.)
👉 And three, because this announcement feeds into the fifth trend I predicted for 2023: the evolution of open-source initiatives that will make Generative AI broadly accessible. I’ve contemplated a future where a few well-resourced actors own and control powerful AI systems. This doesn't necessarily strike me as a good idea. (I covered these themes with Emad Mostaque, the CEO of Stability AI, a few months ago.)
🐺 Yesterday, I spoke with Thomas Wolf, the Co-Founder and CSO of Hugging Face, who predicts that the OS community will flourish, mitigating the emergence of AI monopolies. He also thinks the OS community will offer a rigorous testing ground for key philosophical questions around the ‘safety’ and ‘ethics’ of AI systems. I will bring that to you next week so we can dig deeper.
Now, let’s get to the best of the rest:
The trials and tribulations of ChatGPT, its ‘truths’ and lies
Microsoft’s unhinged Bing Bot, Sydney, has been in secret testing for years. This means the company decided to unveil Sydney with knowledge of its limitations, lies and hallucinations. A slightly bad look for a company positioning as the responsible guy.
Conversely, China is banning ChatGPT because it is fearful of its uncensored responses.
The irony of ironies: in the West, the ChatGPT backlash is led by concerns about what it gets wrong. Meanwhile, the Chinese Communist Party’s problem is about what ChatGPT gets *right*.
US Copyright Office: AI-generated images not copyrightable
The US Copyright Office set a huge legal precedent for the GenAI space by ruling that copyright law does not protect AI-generated images.
The decision was issued in the case of Zarya of the Dawn. A comic created by Kristina Kashtanova, it contains illustrations generated on Midjourney. The creator argued that the images should be copyrightable as her own IP.
The Copyright Office ruled the AI-generated images do not get any of the same IP protections as the text in the comic— because it is “not possible to predict what Midjourney will create ahead of time.”
I’m not a lawyer, but I instinctively disagree with this ruling. I will speak to a leading AI lawyer to explain it in more detail. Watch out for our conversation coming soon.
Spotify launches its GenAI-powered DJ
Spotify is rolling out its GenAI-powered ‘DJ.’ The platform will make music recommendations with an AI-powered DJ to ‘host’ your listening experience.
This is notable as it indicates the hyper-personalization of content : your own version of Desert Island Disks. It’s also interesting as a real-world application that integrates multiple layers of AI — an algorithmic recommendation, a chatbot and an AI-generated voice… in real-time.
🎙️ See my interview here with Alex Serdiuk of Respeecher to learn more about the capabilities of AI voice cloning. Also, watch out for my upcoming interview with Jay LeBoeuf of Descript, in which we more broadly discuss GenAI’s impact on audio production.
GenAI Platform Integrations and PaaS Offering
2023 is the year of mass adoption of GenAI —both for consumers and enterprises. Here are two notable announcements, one for consumers and another for enterprises.
Roblox is building Generative AI into its gaming universe. They’re testing how to use text prompts to allow users to generate 3D models, avatars, code and other creative assets.
Meanwhile, Veritone - a leading AI software and service provider — announced the launch of Veritone Generative AI, its enterprise-grade Platform-as-a-Service (PaaS). It will support integrating and personalizing LLMs (and other generative services like content generation) into organizations.
Open AI, Coca Cola and Bain: GenAI
Open AI doesn’t stop, does it? This week it announced its new strategic partnership with Bain, the management consultancy.
Bain aims to implement Open AI tools (IE DALLE and ChatGPT) to overhaul essential business functions for clients. Coca-Cola will be the first beneficiary of the partnership, getting a facelift on marketing strategy from Bain/Open AI.
I do not doubt that any strategy consultant worth their salt is rushing to understand how GenAI impacts their industry whilst also starting to offer GenAI consulting services.
FYI - it’s notable how quickly Open AI has gone from non-profit to ‘save the world from AGI’ to doing marketing for Coca-Cola. 🙄
Anyway, that’s a wrap for this week.
🎙️ Keep an eye out for my upcoming PIONEER conversations. Of course, share, comment, like, and subscribe.
Our little community is growing quickly, and my mission to make AI accessible for everyone is taking off.
Namaste,
Nina
Happy Friday - was just discussing Hugging Face yesterday. This space is in hyperdrive. I agree with you (mostly) about the IP issue.