AI will change things, but we'll probably still have big tech

One big open question about AI is: How will it change the tech landscape? That’s presuming AI models will have a significant impact, akin to how the iPhone ushered in an era of mobile computing.

One scenario is AI models become a layer in the tech stack, and everybody builds products on top of them. ChatGPT’s recent plugins announcement follows this path. This article looks back at the changes to CPUs to consider the role of LLMs:

LLMs are the new CPUs (Divinations / Nathan Baschez)

LLMs are also similar to RAM in crucial ways. Despite having vast internal complexity, the interface between the LLM and the application it’s integrated into couldn’t be more simple: text in, text out. (Ok, the newest LLMs can take in text and images now, but the interface is still very simple and the point stands.) Unlike CPUs, it’s very easy to swap one LLM out for another. You have to change only one line of code if you use a framework like LangChain. And the cost of using an LLM is still prohibitively high for many applications, so the competitive pressure to get more efficient at scale and bring costs down is very real.

Where it falls short is assuming that OpenAI will lose its competitive edge. Given how far ahead OpenAI’s models are today and the advantage of their Microsoft partnership, there’s a real scenario where a few companies are the dominant AI players in the tech landscape. OpenAI (via Microsoft) might just be a new entrant.

Risk of ‘industrial capture’ looms over AI revolution (Financial Times / Madhumita Murgia)

The phenomenon, which AI experts refer to as “industrial capture”, was quantified in a paper published by researchers from the Massachusetts Institute of Technology in the journal Science earlier this month, calling on policymakers to pay closer attention. Its data is increasingly crucial. Generative AI — the technology underlying the likes of ChatGPT — is being embedded into software used by billions of people, such as Microsoft Office, Google Docs and Gmail. And businesses from law firms to the media and educational institutions are being upended by its introduction.

Unless this week’s open letter to pause giant AI experiments eventuates into anything it seems very likely we’ll still have big tech giants, just with an additional AI infrastructure layer. I remain skeptical open letters cause change. Especially when the likes of Meta, Google and particularly Microsoft/OpenAI have the competitive advantage. They have no incentive to pause.

The Verge poses a much tougher question with AI - where will the money come from?:

Can AI generate a way to pay for itself? (The Verge / Elizabeth Lopatto)

The thing is, both training the model and the specialized people who work with it are ongoing costs. A customer service bot, for example, may need to be fine-tuned every week or couple of weeks. “What’s expensive is that you have to keep doing it, and you have to keep testing the model, and you have to make sure it’s doing what you expect it to do,” Luccioni says. Models also ideally need to be stress-tested to make sure they don’t produce unwanted results. Once all that is done and the model is made available, it may get hundreds or thousands of queries a day. Putting in the engineering aspects that make it scalable and reliable — so it won’t crash — is also expensive, requiring specialized personnel.

Without a clear business model an entire platform shift to AI-first experiences feels hazy but not out of the question given the technical and financial advantage OpenAI / Microsoft have.

Realistically, we’ll likely see LLMs be continually more entrenched in the digital products we use. It’s unclear how much that will change how we interact with our digital devices, but we’ll still likely have a few big tech players dominate our digital lives.

💩 Cool shit

PlayPhrase - Type in a phrase and get a clip from a movie with those exact words

LERF - “Language Embedded Radiance Fields”. Scan a 3D environment and find objects by searching with text descriptions.

ChatShape - I don’t often share paid products but this one is interesting as a signal for where AI-powered products are heading.

Animate 3D buildings based on ambient sounds - I don’t know if this is new? But i just discovered it and it’s cool!

Not By AI - Badges to say your content is human made.

Love at First Line - Pick a book only by a single sentence.

Whichipedia - Which wikipedia page is longer?

Share this email with a friend because We Need Your Confirmation To Ship Your Order