Time to share : ChatGPT: run why not, but know why

In my previous article on the frugality of AI, I had set up the problem of the gigantism of models by simply using current events.??

Well, we have to admit that only a few weeks after this first observation, things are not getting any better... and this is logical.??

It was indeed logical that all the digital giants wanted to stay "in the race" after the opening of ChatGPT, only 4 months ago.?

In the meantime :

Google announced its Bard based on the LaMDA model range.?

Microsoft to forcefully integrate OpenAI from floor to ceiling?

Amazon is rather positioning itself as THE platform to deploy Open source LLMs (Bloom-175B or OPT-30B), but has announced that it also has something in its basket?

Meta is obviously not to be outdone by releasing its LLama and surfing on 2 "well-thought" waves, the first one by indicating that its model is lighter than the others and that it is Open Source. Information relayed without any particular verification by the "specialized" media, even though we are talking about a structure nearly 200 times larger than the famous Bert-Large, which was not reputed to be a small model, and that the model is only usable for non-commercial purposes... so we are far from the Apache license. In short ...?

But in the end, to "stay in the race" of ... what???

Because we are facing a rather new case with these models and it is clearly OpenAI that, by making a usable version of GPT3 by the general public, has caused this situation.??

Indeed, by making a version of GPT3 "usable" by everyone, the general public discovered what had existed for a long time. And it seized this new digital toy in a dazzling way, here by inventing new uses that are increasingly redundant and often of dubious usefulness and reliability, there by having to prophecy and share cookie-cutter visions of new businesses on the cusp of disruption (it's been too long since we've been able to use that word, that must be the explanation??). And in many cases, this has managed a lot of self-proclaimed experts giving advice on technos they have absolutely no command of.??

The fact is, the real uses are not yet clear, the reliability is not yet there and none of these systems are optimised for mass use!??

No, the technology definitely does not make the use! (This is the 2nd credo of our Innovation Department at Inetum!)?

And yet it's a race!?

We have to make the APIs available, we have to put Bard online, we have to integrate chatgpt to our chatbot service (and while we're at it, we'll put Dall-e too, we don't really know what it's going to be used for, but at the very least, "it looks good, it looks innovative").?

?

And when you see everyone running around you, well you inevitably tend to ... start running too!?

Without finally knowing why and especially without knowing where you are going!??

And yet, there have been many cases where we have rushed into the technology without really thinking about the usefulness, this is unfortunately far from being new!??

How many IoT POCs have been left in the wardrobe???

How many blockchain prototypes were made before it became clear that a simple electronic safe would be more than enough to cover the need???

Not to mention the recent and rather bitter failures of the largest metavers initiatives...??

?

All these failures are due either to a clear lack of hindsight and a little functional challenging before launching the projects, or sometimes to neo-experts who take advantage of the technical credulity of the business departments, surfing on the undeniable whaou effects and who, once the budget has been burnt up (if you'll pardon the expression), leave the client in his disappointment and disappear behind a smoke screen.??

What you have to remember is that fortunately there are many great stories about all the technologies I mentioned (we have many to share!!), and this will inevitably be the case with ChatGPT (well no, probably never with ChatGPT as such, but let's say with the underlying technologies), but it's really time to take a step back, to sit down and think calmly about the real objectives, the real usefulness of the use.??

?

All the more so since, and I'm coming full circle to the point about frugality, all this "digital agitation", these junk services, all these cases of uselessness, are VERY expensive in energy...??

? Jean-Paul Muller Global Practice Manager IA Inetum