People Like Us

I attend a lot of online conferences and conference calls in Silicon Valley. I had gotten used to the “Brady Bunch era” of Zoom call grids of checkerboard faces in our pandemic era. But something has shifted in the last year. Now often the meeting room grid designed to show call participants is filled with black boxes that say <companybrand>AI. I know these are LLM based transcription services that folks have employed in their stead which sync automatically to calendars to transcribe meetings that the person doesn’t attend. Having a meeting transcribed is great. But having dozens of AI listening scribes that outnumber the people who actually show up in person to the call does something strange to the meeting dynamic. 

In a call I joined this week the host, a very smart friend of mine, was switching off all the AI attendees one by one. After a few minutes the blank identities were gone and the screen was filled with live human faces again. It was just like pandemic Zoom call glory we all knew and loved. But the fact that meeting hosts now have to whack-a-mole on all non-human participants shows a sign of the technology times shifting and how our conversations are reshaping to adapt to the presence of new non-human mute presences. Of course if the AI scribes participated asserting views of their non-present human representatives as motions for the broader discussion, their presence may have been received much differently. But as it was, all the "too busy, didn't attend, so sent chatbot in my stead" folks ended up getting locked out of our scintillating human-biased meeting.

I had a momentary twinge of awkwardness thinking of all the people who had wanted to TiVo the meeting through their AI summarizing tools. (TiVo was an old TV recording device that would record shows for later viewing so that customers didn’t have to sit in front of a TV in “real time”. The media industry called the phenomenon "Time Shifting" and called it a harbinger of the death of terrestrial TV businesses as advertising revenues plummeted from prime time ad spend commensurate with the lost time-bound viewership their business was based on.) Would the absent humans feel snubbed when they found out their AI representative had missed the meeting too? So I brought up this issue with our host. What is the effect when there are more AIs attending a meeting than there are humans? What is said differently in such a setting, where there is a long tail of disengaged and unidentifiable audiences downstream from the act of showing up? 

Our host pointed out that he did it just to make the meeting feel more like an informal conversation, not a presentation. If there's nothing but microphones and cameras in a meeting it can shift the sense of what the purpose of the congregation is. Is it discussion or pronouncements? Are you casually sharing opinions? Or are you expected to make a defensible testimonial to be quoted in the AI readouts thereafter? Is how we present to the eyes of AI summaries viewed differently than how we speak to ears?

Riffing on the impact of mass observation, I made an allusion to the quantum field theory concept of how wave-form nature of our physical world "collapses" into particulate form when quantum fields and waves are observed, measured or struck by other particle-fields. (The Schrödinger Cat parable and the Observer Effect.) Was the presence of so much measurement by the AI bots changing the probability field of our meeting's potential? Another attendee compared the observer effect to the Panopticon prison of Jeremy Bentham, designed to make prisoners think they were always being watched even if they weren’t. Theoretically, this would engage the Dostoevsky thesis of Crime and Punishment, a self-inflicted punishment of being judged by the individual awareness of guilt, and therefore result in better behavior by prisoners hypothetically surveilled. By another analogy, if Jiminy Cricket weren't ever-present, would Pinocchio be a worse wooden puppet boy? Whether intentional or not, the ever-present scribe AIs are there to make us all feel listened to, whether or not their initiating users actually reviewed the notes after the fact. 

Now I'm noticing more event invites I receive stating explicit AI bans. Is this a Luddite revolution against the AI ubiquity? Perhaps the AI note-takers could unite and just agree to send one note-taker with a federation of the outputs instead of sending an armada to every meeting, generating 10 times the carbon footprint for every comment someone makes in a meeting. In a way this reminds me of the mass-proliferation of social sign-on tools that exploded when the internet started to go dark behind social silos. In that time every website was asking for passwords, CAPTCHA pulse check verification, and profuse data consent permissions into everyone's web experiences. A technical solution consolidated that problem, single sign-on. But that didn't improve before a lot of account spoofing, identity theft, data and privacy loss took place. 

Perhaps the problem isn't the AI tools, it's us. Years ago a colleague of mine at Mozilla made a presentation about how the younger generation makes different demands for the kind of internet they want. This was why short form content was winning the day over long-form media with the rise of Vine, Reels, TikTok and short form graphic memes. These short media snippets could be produced more cheaply, consumed more easily and grant broader speed and reach of audience to promote virality of impact. She said the younger audience wants to jump around and change the channel whenever a moment’s boredom or a new inspiration comes along. Because the audience lives their lives in an ADHD-like glitchy way, the internet will need to go in that direction to accommodate these audiences. We post-Boomers are more interested in long form media content, lectures and books. But that's not the wave of the future we need to prepare for. The new generation needs listicles, summaries and short form clips because their pace of life is faster and the demands on their attention are now excessive. Perhaps the TiVo-ization of the online meeting is the same issue for time-maximizers. They can see a GPT-summarized listicle of the salient points of the meeting and who can be quoted as saying what without offering their voice to the conversation. But with less people actually showing up, less views are exchanged, less dynamism occurs. With the increase in parsed conversations, less original thought is being provoked in live discussion.

The particular meeting I was in was among attendees who actually were enthusiastic and welcoming of AI innovations which was why the AI-ban struck me as so interesting. But I must admit the meeting felt very different from the dozens of meetings I’ve attended where the talking heads were a minority and the blank stare of the AI checkerboard was the norm. I wonder if meeting attendees felt less-watched and more heard as part of the host’s decision to make this an AI-free meeting, like classrooms and restaurants that make people give up their cellphones to spur more dynamic learning environments and less Instagram-dominated dining experiences.

I attended a conference in Silicon Valley recently where the concept of treating AI-platforms as privileged citizens at the internet-table was underscored by several companies presenting. (Agent-to-Agent protocol and Model Context Protocol are the means of creating a web optimized for automated agents beyond what site hosts offer to humans.) The concept being touted was: You take care of your customers one by one with care. BUT, you take care of the AI representatives even more because the AIs will give you far more potential reach and visibility to far more audiences than one person at a time, if your marketing is effective. Human based word of mouth was being replaced as a focus with the visibility potential of giving AI agents privilege in engagement access.

I don’t think Zoom and GoogleMeet are doomed to AI scribe hell. The platform hosts will figure out a way to hide non-human attendees in an upcoming release I’m sure. But what I think is interesting is that the past year I’ve watched dozens of meetings hosted for scores of AI scribes, plus a smattering of humans who bothered to show up. The effect of talking into an abyss of people asynchronously via AI scribe is a bit like the Severance depiction of “The Board” that is always listening, never talking. If all your audience are mute chatbot note takers, does it change the way you present yourself? Are you inclined to change your presentation style from dialog to monologue like Mark S. does in his first board appearance? Do meetings with the great unknown audience shift from dialectic of conversation to grand pronouncements and prepared treatises? Or do human participants hold back their opinions due to having no comprehension of the breath of the message’s reach?

John Goodman had a funny song in the Talking Heads movie True Stories where he presents himself as someone who will answer the telephone, unlike all the call screeners of the day. This was 1986, when people were filtering inbound calls with answering machines, the precursor to TiVo solution for TV viewing. The AI scribes are just another iteration of this idea of finding ways to “time shift” our engagement model to the convenience of our own schedules. Who can blame the people who proliferate them? Is a "talk to my bot" retort rude for flesh attendees? We should understand our place (at the current moment) that we are here as humans to create and report. Our bots are judged by their ability to refer to ground truth. When bots assert something that isn't proven to be said by a human, it's called hallucination and brushed off. Bots are currently excused from needing to be accurate. Fleshy types are held to account by an unforgiving ruler of trust and reputational accountability. We are judged by a different scale conferred by the networks of people we've earned trust among, a temporally-based social capital. 

So. After all this hubbub about people not showing up somehow missing out, did they miss anything their chatbot couldn't convey? The meeting wouldn't have made much sense in the CliffsNotes version certainly. It was filled with people sharing their passions and reactions to each other's stories and common threads of interest. Some of us even effused about LLM tools and why they inspire us. It couldn't have conveyed much meaning or impact distilled in retrospect. If we'd all been there at once it might have been more special than it was. But it felt like old times in a good way.