chatgpt | Successful Software

Recently a customer emailed me:

Please consider re adding the option to convert from one time zone to another as was available in [Easy Data Transform] version 1.x. See the attached screen dump.

Easy Data Transform has never had a time zone conversion feature, so that is a bit strange. Although . But the screenshot really set off alarms bells as, despite saying “Easy Data Transform v1.11.2” in the title bar, that is not our software!

I thought that someone was trying to pass of their product as ours and did a search. But I couldn’t find any reference to another piece of software called “Easy Data Transform”. I emailed the customer to ask where they had got the software from. The reply came back:

The screenshot was from a ChatGPT output – maybe hallucinating.

Wow. Not only had ChatGPT hallucinated the feature, but also a fever dream screenshot of the user interface, with the non-existent feature. It looks like a real screenshot at first glance, but the icons are a giveaway if you look closer.

For reference, this is what the actual user interface looks like:

ChatGPT has got some of the input types, transforms and menus correct. But otherwise it looks quite different.

I had another brush with AI hallucinations when I asked MS Copilot how to perform ‘one hot encoding’ in Easy Data Transform. It came back with a very plausible and confident sounding answer, including this summary of the transforms required:

Just one problem – only 1 of these 4 transforms actually exists (Split Col). It hallucinated the other 3!

Customers are increasingly typing questions into AIs, rather than reading documentation or asking on a forum. That is good news if it means that the customer gets a quick and accurate answer without troubling busy developers. But it is very bad news if they are getting incorrect answers, especially when these answers look plausible and are confidently presented. It is galling enough that AIs are stealing all our web traffic, without them giving our customers bad support advice as well! Wild times are ahead.

Ps/ Time zone conversion and one hot encoding are now available in the latest version of Easy Data Transform.

I’m sure we are all familiar with the idea of a technological singularity. Humans create an AI that is smart enough to create an even smarter successor. That successor then creates an even smarter successor. The process accelerates through a positive feedback loop, until we reach a technological singularity, where puny human intelligence is quickly left far behind.

Some people seem to think that Large Language Models could be the start of this process. We train the LLMs on vast corpuses of human knowledge. The LLMs then help humans create new knowledge, which is then used to train the next generation of LLMs. Singularity, here we come!

But I don’t think so. Human nature being what it is, LLMs are inevitably going to be used to churn out vast amount of low quality ‘content’ for SEO and other commercial purposes. LLM nature being what it is, a lot of this content is going to be hallucinated. In otherwords, bullshit. Given that LLMs can generate content vastly faster than humans can, we could quickly end up with an Internet that is mostly bullshit. Which will then be used to train the next generation of LLM. We will eventually reach a bullshit singularlity, where it is almost impossible to work out whether anything on the Internet is true. Enshittification at scale. Well done us.