A Note on Not Being Shocking, and Making Connections

I don’t often do personal updates, I just sprinkle them in, on the off chance anyone wants a bit more behind-the-scenes. Two things come to mind to mention today: the repercussions of not being shocking and of meeting AGI Insiders.

First, is it me or has AI coverage devolved a fair bit more in recent months, across the board? The way algorithms incentivise sensationalism is brutal, and clear, especially as a content creator myself. And trust me, ‘SHOCKING’ works, and not just for views - it brings some people life-changing numbers of social media followers.

Here’s the playbook on Twitter:

‘Claude 3 just ended THESE seven careers, and the wild thing is, only a few people really understand what’s coming’. Then 3-5 tweet-replies, that don’t actually support the argument if you dig just a bit, then a link to an account/course/channel. Or 'HUGE news' with a link to an off-the-cuff Gates/Musk quote...

And of course, on YouTube, just claim Everything stuns Every industry, shockingly [GPT-8]. Plays into deep-rooted and genuine fears of displacement from AI, so gets insane views. The other problem though, is it fills the airwaves with static, so the busy public struggles to discern when really transformative things happen, as they sometimes actually do.

The real world of AI developments and career upheaval is important enough not to wildly exaggerate and hype on a daily basis.

The clear ethical dividing-line, for me, is the question ‘Does the headline faithfully convey the content and the beliefs of the author?’.

So take my two boldest titles in 2024: ‘Nukes in Space?’ and ‘The New, Smartest AI’. Definitely eye-catching titles, but did they match the research and my beliefs? The first was based on reports in Reuters, Bloomberg and the NYT, the second on a suite of benchmarks and personal tests. I can honestly say I have never written a title or tweet that I didn’t believe, to the word.

My message is this: be aware of these deep, powerful incentives when watching all creators, even me. Hold us accountable, with your clicks and support, if people say or write things that just aren’t true. I don’t mean mistakes. I mean, well, lies. Is ‘everyone’ ‘shocked’ or ‘stunned’ by this new model? Is this feature of GPT-5 actually ‘CONFIRMED’? Well, chances are, the author knows they are not.

Second update: I am quite often these days chatting and physically meeting up with those who work at OpenAI, Microsoft, Google DeepMind, Meta, et al. And of course maybe 20% of you reading this work at these places yourself!

But there is a hidden responsibility with having these personal links that I wanted to voice, partly to thereby hold myself accountable.

My obligation is to communicate as clearly and honestly as I can, first to the public, and also, of course, to supporters. The more you meet people whose work/organisations you may end up criticising, the more tempting it could be to tamp things down; avoid pissing off people you know will be watching. Become more ‘industry-friendly’ to secure bigger interviews, get earlier model access perhaps, or good gossip. I believe I have so far resisted this pretty unambiguously but if you ever see me prioritising the maintenance of such connections over impartial coverage, do give me a little slap!

So, yeah, that’s the update. I know I can do better too - deeper research, more clarity - and am not speaking from ‘on high’. But the above might be interesting to some, to see some of the incentives at play for those covering AI.

We must all watch for those ‘insanity points’ that Sam Altman warned about, as the world strides ever closer to AGI.