How I learned to stop worrying and love the bomb (at least a little) | beeb6502.com
Open navigation
Introduction
In this piece, I present several reasons for being an AI sceptic before making an argument in favour of the use of LLMs in software development. I intend to show that while AI scepticism is valid, it should not crowd out the genuinely powerful capabilities that generative AI brings to the software development domain; capabilities that I, too, am now adopting, albeit with a number of significant guardrails that I will explore in more detail below.
First, some home truths
When<br>ChatGPT first launched
, back in late 2022, I was not its greatest fan. Perhaps because of my proximity to the world of machine learning as a data engineer of some 20 years standing, I felt that a lot of the gush that accompanied its release was due, in large part, to the general public finally getting a glimpse of what generative pre-trained transformers can do and believing it was some kind of miracle. It was not, and I remain steadfast in my opinion that this remains true. You don’t have to be<br>Arthur C. Clarke
to know that scientific advancement, especially at speed and scale, is still science, even if some of its consumers fail to distinguish it from magic.
Large language models (LLMs) are impressive. But like any technology they have both limits on what they can do and flaws that only become more apparent the further they are pressed to operate outside of those limits. For all that LLMs have advanced over the last several years, I do not believe any of that has changed, even if their abilities, along with the number of domains in which they are able to operate, has significantly increased. When used with sufficient care, attention to detail and safeguarding, they are able to carry out vastly complex operations with only natural language as an input. Anyone who claims that this is not true is, in my opinion, choosing to look the other way. As it happens, I have a lot of sympathy with that position. It isn’t difficult, after all, to find reasons for disliking generative AI.
Why people hate generative AI
First, the techbro billionaires behind the generative AI boom of the last four years are, to a man (and they are all men), very much the kind of people we’d like to strap to one of their beloved rockets and fire off into space. From Altman to Zuckerberg and all the others in between, they are living proof that money, even in vast quantities, cannot buy basic humanity. It is something of an irony that these same techbros like nothing more than to launch a product with the words “imagine if you could…” only to then follow it up with something<br>monumentally tone deaf
or<br>bone-headed
. I cannot remember a previous technology whose leadership was so united in its moral bankruptcy, even when judged by the admittedly flawed Silicon Valley standards that gave us NFTs, the metaverse and Theranos. That the same people making wild claims about GenAI on LinkedIn were, only a few years prior, running their own crypto-scams is no coincidence. And wanting to have nothing to do with these people is not only understandable, it is righteous.
Second, the<br>environmental concern
that accompanies the build out of data centres and every single generative AI query sent to them for processing neither can nor should be ignored. The energy and potable water consumption is staggering, and carries an even greater cost for the communities upon which these data centres are inflicted. The<br>grassroots protest groups
that have sprung up to push back on local authorities for authorising the build-out of such climate-change-maxxing projects should be commended. There really is no satisfactory answer to the impact GenAI is having on the planet, and the fact the big tech firms are being allowed to get away with this is a criminal failure of government.
Third, there is the hype. The remorseless, 24/7 hype that is blowing generative AI out of all proportion. It is very easy to see why some people hate it when, day after day, they’re told it’s<br>going to take their jobs
. Not the jobs we’d be happy to wave goodbye to, but practically every kind of human creative endeavour we’ve got: music, books, the arts, cinema, practically anything we do for our own entertainment is apparently under threat from AI, if the relentless social media hucksters are to be believed. And knowing that this is only made possible because the frontier AI labs were given carte blanche to hoover up every piece of creative output we’ve ever produced as a species only serves to rub salt into an already weeping wound. Technology was supposed to set us free to pursue our own interests; but the hype train is pushing a message that these, too, are going to be taken away by machines. What makes the hype more infuriating is just how much of it is premised upon what AI “might” do in the future. Have you ever looked at...