Saturday Links: Pins, Agents, and our Digital Souls

Enjoying the last few days of warmth as we reach what seems to be the tail end of summer in Northern Europe. Here are this week's links:

• Techcrunch Review for the upcoming Plaud Note Pin. So far specific AI hardware hasn't been wildly successful (apart from maybe Ray Ban Meta Smart glasses which I love and I still have hopes for the Rabbit R1 form factor). Things will get better though, the Plaud note pin might be one such device: focused just on note taking it clips to your lapel or can be placed on a table and has a well optimized pipeline for transcription and summarization using ChatGPT. I do think this kind of device will make sense for when you don't need a screen around (which I hope will increase). I would worry for the company though that the hardware simply becomes commoditized with such a narrow use-case and the value ends up in the transcription and note management, which in turns ends up absorbed in a broader note/personal data management stack.
• Salesforce announces Agentforce and $2 per customer conversation pricing plan. At the Salesforce flagship dreamforce conference this week, CEO Marc Benioff annouced the company's new product offerings that make it easier for companies to build sales and support agents to interact with customers. The argument Salesforce makes is that they are best place to make it easy to create excellent customer experiences (in part because they hold and manage a great deal of customer relevant data in their platforms and because these experiences are hard to create). Certainly Salesforce will compete in the agent creation markets, but what's perhaps most interesting is the pricing model: per agent-customer conversation starting around $2. Press coverage is focused on the potential message from Salesforce that AI will inevitably cost jobs in sales, marketing and customer success. I think this is clearly true. The bigger question here though is whether a shift to creating an "agentforce" will enable a company like Salesforce to keep growing. It's likely downward pressure on their high priced per-seat model could quickly outstrip agent based revenue. The main advantages that salesforce has is all that customer sales and support data. My suspicion is that moves like this are more about holding onto that for as long as possible to retain the status quo than spinning up new revenue.
• AI doesn't "write". This one isn't a link, but an opinion. In response to another article out on the Internet, a good friend of mine wrote on her blog this week something to the effect of "we should stop with calling what AI does "writing.". I think it's an understandable point of view, but I have to say I rather disagree, just not in the obvious way. Yes ... today's LLMs type chat systems are token generators and create a probablistic stream based on some inputs. But, I think as we learn more, we'll learn that in many ways that's exactly what we humans are doing. We're a collection of knowledge, experiences, emotions and so on and when we're prompted with something we react - we "write" (or speak, shout, run away, ...). It's a lot more sophisticated and nuanced than what an LLM does, but I suspect that we'll find it's fundamentally the same thing. That sounds pretty horrifing if you take if to mean "we have no intent" or "we have no soul", but more optimistically I like to think our memories, emotions, experiences + some initial basic programming etc. really are the thing that is our soul. (... and thank you to the writer for this prompt!)
• OpenAI closes in on largest VC round of all time. $6.5B at a $150B valuation reportedly. If this was a public company it would edging into the top 100 companies globably by Market Cap. Something tells me, though, that this wont stay a fundraising record for long. AI will create more funding monsters yet. I suspect for the investors in this round, it'll also be around being at the epicentre of the transformations openAI will bring and understanding their impact for other investments. That's probably worth a very large premium.
• Mark Zuckerberg says Europe needs more consistent AI regulation. A new critique of EU Privacy and AI regulation from a range of major companies includes multiple European companies as well as AI giants like Meta. The ask is for clarity and consistency, but underneath there's also an ask to enable innovation. I don't know the exact place to draw the line, but I fundamentally agree with what the companies are saying. First and foremost there should be clarity. Second, in my opinion, people outght to be able to opt-in, Third, there is a danger of European culture not being represented in global AI models. Fourth (and most important of all in my opinion), AI systems have already become some of the most powerful productivity tools on the planet. Denying or reducing the ability of EU companies to participate in building them and EU citizens the ability to use them risks leaving European behind.

Lastly, it turns out you don't need AI to fake things. This Chinese Zoo just admited it's pandas were in fact disguised dogs. The dogs were presumably happy to be released from their bamboo diet!