ChatGPT integration free trial, Bullshit, Botshit and Bubbles (newsletter 3)

ChatGPT integration: join the free trial

Later than promised, but it's here: How to chat with ChatGPT about your content.

As an early adopter you can join the free trial: simply apply by filling in this form, and I'll add you to the trial. The trial will continue until the allocated budget has been exhausted.

Latest experiments

(Newcomers: you can find all my experiments into sending Hubbed notes to ChatGPT summarised in posts tagged #myhub AND #chatgpt AND #experiment, and shared in full on https://experiments.myhub.ai/.)

Since my last newsletter, which focused on "visualising knowledge" (experiments 2-4), I've been combining and applying the various prompts I've tested on content as diverse as excerpts from a LinkedIn Live interview I did with PR/Comms company ZN and the first 5 resources I curated this year. Along the way I've had ChatGPT:

• try to integrate internal and external communications; 
• conceive three EU-sponsored online communities (organic agriculture Garden, GreenFieldsEU and EU Converge) 
• write 5 blogposts on what Open Source offers AI, in the process comparing GPT 3.5 with 4 (see experiment 10).

What did I find? Read on...

Bullshit, Botshit and Bubbles 

I was relieved to see I assure you, an AI didn’t write a terrible “George Carlin” routine from Cory Doctorow recently.

Cory introduces a new word I'm pretty sure we'll all be hearing a lot of this year: 

"there’s a huge difference between producing a plausible sentence and a good one. After the initial rush of astonishment, the stench of botshit becomes unmistakable".

From my notes: the ubiquity of botshit stems from the fact that so many people benefit from it - not just the tech's purveyors (obviously), but also:

• critics: "doomsayers who form an unholy alliance with AI companies by parroting their silliest hype in warning form"
• users who, "when the bot fails [them]... manufacture their own botshit, assuming they must be bad at prompting"
• and people like me, who are integrating these tools into their products and trying to hype them as they "quietly hire legions of humans to pick up the botshit it leaves behind"

Except that's not what I'm doing. Yes, I'm exploring the How and Why of integrating ChatGPT into MyHub.ai. But, to answer the question in the previous section, I've yet to find a truly convincing, knock-your-pants-off application. I'm still looking, and I'm opening the free trial to see if anyone else can find one, but I'm not selling botshit to anyone.

Because that's how hype cycles work: there's a splash of excitement; a surge of content as everyone hops on the bandwagon to build followers and sell books and courses; a plunge of disenchantment when the Easy Wins promised don't materialise; ... and then, eventually, those who stuck with it without riding that rollercoaster figure out where the actual value lies.

If, indeed, there is any. Which brings me to ...

Bubble Trouble

... another post by Cory: What kind of bubble is AI? From my notes:

• The cost of each AI query includes "a massive primary energy bill... [another] for the chillers, and a titanic wage bill for the specialized technical staff. Once investor subsidies dry up, will real-world ... applications cover these running costs?"
• No, because most of the applications driving the high valuations we see are high-value but only if they're risk-tolerant. However, the "risk-tolerant applications are almost all low-value; while nearly all the high-value applications are risk-intolerant". Uh-oh
• So what will be left? "smaller models... run on commodity hardware... [with] communities ... formed around them", pushing their limits "far beyond their original manufacturers’ beliefs about their capacity".

If you dig into experiment 10 you'll find me asking myself something similar:

"while the GPT4 outputs were clearly better, I'm not sure they are 100x better, while GPT4 costs 100x more." - experiment 10

So while I'll continue these experiments, Cory's last point about open source (above) reinforces my original conviction for MyHub (see How Artificial Intelligence will finance Collective Intelligence, which needs an update). Expect to see more on open-source / local AI in the future, perhaps using GPT4's blog post about it as a starting point.

There's plenty more bullshit where that came from

There are currently two more articles in my Hub tagged #AI and #Bullshit, while if you look for just the bullshit you'll find 15, covering everything from bullshit jobs, bullshit productivity tricks to bullshit Presidents and Prime Ministers.

It's just one word, but it explains a lot. Which is why I suspect those numbers will grow in 2024.