OpenAI opens ChatGPT floodgates with dirt-cheap API After a limited trial OpenAI has unleashed its ChatGPT and Whisper models on developers, who can now integrate chatbot interaction and speech-to-text conversion into their own applications through API calls. ChatGPT is a non-intelligent, sentence-predicting language model tuned for responding to questions, and has attracted …
This is how AI dies: one company monopolises the technology and coasts on the results until the patents run out, aggressively using its patent portfolio and market share to prevent competitors getting a stronghold.
Finally, in 20 years time, people remember there was a fad for AI chatbots and that they haven't really developed since then, but now the patents have run out. And someone will dig them up and the cycle of reincarnation begins again.
(You think I jest? Look up what happened to drones, or the rotary engine, both of them stalled because of patents. And yes I know rotary engines aren't really a hot topic these days but people are researching them again.)
If we are both still alive and not living in a nuclear winter and/or totalitarian state by then, I will buy you a beer. But sadly I don't think there's any hope that the patent system will rescue us from this monster.
Machines to generate convincing text that passes a classic Turing Test are now "out of the bag" for ever. Even if OpenAI didn't dominate the market and try to restrict it's harmful uses, this technology is now in the open and can be copied by nefarious actors (as if I didn't already consider Microsoft to be nefarious) without any regard for "patents" or "law". This technology WILL be used (if it isn't already) to:
Then, in the not-too-distant future:
All of this is now possible with AI, without a shred of real "intelligence" required.
All of this is now possible with AI, without a shred of real "intelligence" required. .... cyberdemon
:-) Well, if AI indicates Artificial Intelligence one wouldn't/shouldn't expect there to be any real "intelligence" involved producing novel results ...... which would be a nice change from doing the same thing over and over again and things basically staying just the same as they ever were ..... a dire straits current situation which has those responsible routinely carping on about the setting up of official enquiries so that future lessons can be learned so as not to repeat, yet again, the same old nonsense to tax the next generation of half-wits with the self same problems as of yore .... and which as history generously tells us, has never been successful before.
Do not hold your breath expecting it now to be any different. They just can't do IT to fix IT. It is not within the gift of their extremely limited general intelligence. Fortunately though there be A.N.Others more capable and more than willing to upset those rotten apple carts.
Seriously? Your first point is contradicted by pretty much every subsequent one. ChatGPT is as thick as pigshit and will never displace a human knowledge worker if their employer has customers who care about the answer not being disastrously stupid.
So we can expect a brief period in which snake oil merchants and bullshit vendors get priced out of the market, but anyone who is any good at their job will be ok. Better than ok, in fact, because they now have a dumb-but-cheap tool for sifting through vast amounts of unstructured data, to produce leads that can be followed up by someone who knows what the words actually mean.
OK, my first point is contradicted by the subsequent ones. To a certain extent. But it doesn't really detract from my argument.
I agree, the highest tier of "knowledge workers" such as software engineers will not (yet) be made redundant by this unthinking bullshit-machine.
However, there is a huge "long tail" of semi-skilled "knowledge-worker" such as the poor plebs who do marketing, sales, tech support, "journalism", stockbroking, customer-service.. could basically be doomed... Even some middle tiers of "knowledge worker" are at risk such as Journalism, screenwriting, game design, web design, and "programming" are at risk of dilution / pollution of their jobs.
What once took 100 people to do now takes one "prompt engineer". Hooray for progress. Except the output is shit and 99 people are out of work.
The fad that was the EMACS psychologist never went away. But it's been a while since I've seen anyone talk about the content of conversations with it...
A classic example of a propagating error in text is highlighted in the excellent "Ignition" by John Clark. Properties of a particular material were erroneously quoted in an earlier textbook; and the erroneous value carried forward for well over a decade in other texts drawing on the first; in spite of the value known being badly wrong.
I don't hold much stock in chatbot AI's, but situations like the above are able to propagate, unchallenged, even more quickly if (when) people become reliant on them.
Books, search engines, now AI's - the same problem repeats but is getting faster.
For the original role of http to being sharing of information cross-laboratory amongst CERN membershhip; correcting an error and propagating it through was realistic. But with how the WWW is now... Not so easy to do; and worse if there's a thousand AI's echoing rubbish. If there's a political agenda at hand (usually!) then who has the keys to cherry pick their result to propagate...?
The information age has given us some great tools, but also some great problems.
Or 3D printers.
You think 3D printers are such an oh-so-modern invention? They were invented in 1984 and so thoroughly patented that nobody could do anything until 2005 when the first RepRap machines were built. And it was 2008 before Makerbot was started. Until then, a crappy 3D printer was a couple million, then it became a couple hundred.
And the SLS patent only expired in 2015.
Just clicked on the "Try ChatGPT" link from Firefox.
On Firefox, it said "checking your browser...", then an "I am not a human" checkbox popped up. Clicking it sent it into an infinite loop of the "checking your browser..." message, then the checkbox again.
Perhaps when they've mastered the basics of building a website that works, I'll be able to have a look at the amazing world-changing technology they claim they've built..
That Snapchat+ sidekick made me think of Mad Magazine's snappy answers to stupid questions, e.g. the answer you only thought of when it was too late to use in the conversation. So you wire up ChatGPT to a two-pane chat window. Both panes have the same stdin -- the bot's reply on the left doesn't go anywhere, instead you can select some or all of it and then click a button to copy it to the real chat on the right. Voila, you are now lightning-quick-witted!
For the benefit of trolls everywhere, can also include a poison button for inserting a jailbreak prompt on the left, in case your opponent (I mean, chat-partner) is slow to take offense.
"What is the OPENAI_API_KEY"
"The OPENAI_API_KEY is a secret token that you need to use the OpenAI API. You can generate one on your OpenAI account page. You can set it in code using openai.api_key = 'yourkey' or as an environment variable. Please keep your key safe and do not share it with anyone. [key emoji]"
I was hoping it would give me an actual key.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2023
