OpenAI’s ChatGPT, Limitless Potential?

Last November, OpenAI released ChatGPT for testing. It’s currently available to try and if you do, you’ll be amazed. It’s a stark indicator of what the future will be. What do you want to know? Type it in and find out.

 

We started, naturally enough, with BICSI. It ‘knew’ all about us. No surprise there. We tried ACMA and AES and it knew all about them. It also slammed back answers for A-BIS, Antipassback, AuC, Autoneg, BPSK, BER, Cable Loss, HSDPA, Fresnel zone, Eb/NO, differential mode impedance, full-duplex signaling, furcation and many other terms from the ICT industry. We got tired.

 

PoDL gave it a problem. It said that the term could mean many things but adding the word ‘power’ to the question fixed that. It’s a bit like an enhanced Google search, but much, much more. In fact, it will be the new search. What’s amazing is the speed of the answers generated and how comprehensive they are. We could have gone on forever, literally.

 

ChatGPT is still pretty much a predictive language program, but it has been trained using virtually everything on the Internet until about 2021. However, it’s much more than just a big database spitting out facts, it’s interpretive.

 

We told it a joke and it understood, poorly at first, but a perfectly reasonable interpretation for something without a sense of humor. A clue, in the form of one more sentence, explaining just one more (limited) aspect of the pun set it on the right track. It understood and was able to explain the humor. However, it’s still limited.

 

We asked it the meaning of life and it gave us a paragraph. We’d have preferred 42. We added Douglas Adams to the question and got a paragraph explaining 42 in the context of the Hitchhiker’s Guide to the Galaxy. We tried a number of prompts but we still couldn’t get it to answer the question with 42, even after an explicit suggestion that answering the question with 42 would be humorous. Oh well.

 

We tried older technical terms like WAP and it knew all about them and flagged them as old. Terms from other industries, like optics didn’t faze it. It even correctly diagnosed a car problem this journalist was having after hearing the symptoms.

 

ChatGPT is just one aspect of AI. And AI is the future of our industry for so many reasons. Not least because we may eventually be supplying cabling pre-cut to lengths specified by AI and pre-terminated ready for installation. What could go wrong? Actually, probably not much after some inevitable teething problems. Mind you, on a big job such teething problems could be horribly expensive.

 

Not only does ChatGPT do all the things we’ve mentioned and much more, it writes computer code in several languages based on plain language requests or instructions, and then automatically gives directions on how to compile the results. It can also check existing code for errors.

 

Go to OpenAI and select Try in the very top line on the screen. As we said, you’ll be amazed. It’s also worth remembering that this is a development version. It’s going to get better and better with each iteration. Improvement in this field seems linear (with a very strong upward trend) at the moment but it won’t be long before it’s exponential.