by Jo Nova
Another Great Disrupter?
What if you had access to an experimental Chatbot that could write your reports, debug your code, design your new ad campaign and answer all your questions? We are already using Google or some search engine to find links to these answers. But now an AI has been released that may disrupt all of that and a lot more. ChatGPT was trained on text culled from the internet and it creates the answer live in seconds — it writes it out in a conversation with you, and in Python code, Norwegian, iambic pentameter, whatever you want. It can do birthday suggestions, business plans, eulogies, speeches — like a personal assistant with a copy of the entire World Wide Web in their temporal cortex.
Google, the dominant gatekeeper to the internet for 20 years, suddenly faces a “make or break” point. ChatGPT could wipe out its’ business model and Google has issued a “Code Red”. An extinction level event couldn’t happen to a nicer company, but will ChatGPT be better?
It was launched on November 30th, and is taking off. Jorden Peterson asks, is this Gutenberg Press Level? It’s engaging, wow.
…
Right now, it doesn’t seem like the next generation Gutenberg breakthrough. It is surprisingly good with human sentences and appears to be wowing humanities professors. But in my short trial asking science questions rather than humanities ones, the conversation became like a self generating Wikipedia thread, because the internet is what it is today, and the AI “knows” boring politically correct things. Having said that, if I want IPCC type answers, this was much faster than Google, and faster than wading through IPCC Chapters. So it is a threat to Google, but not a threat to the IPCC. It is not speaking truth to power.
That may change in the next year as the AI learns to dig through the dark corners of the internet and to resolve some very difficult questions. But how much is it “guided” from behind. I don’t know. There is plenty to worry about. If skeptics keep asking it for evidence, and drilling down into those details will the AI learn to join the dots, or will it merely repeat IPCC permitted reports, and be even better at hiding skeptical answers than Google is?
I asked ChatGPT if it was open source, and it replied:
While the underlying model architecture and training techniques for GPT-3 are publicly available, the specific model parameters and training data used to create GPT-3 are not open source. As such, it is not possible to access the full source code for ChatGPT. However, you can use the API to interact with the model and generate text based on your input.
So, whoever controls the model parameters and the training data could be the new gatekeeper of the WWW? What if the model parameter says the UN is always right?
It was co-founded by Elon, but Microsoft have bought in:
Made by OpenAI, a San Francisco-headquartered AI research lab co-founded by Elon Musk, ChatGPT is capable of understanding natural human language and generating thoughtful human-like prose after being fed a prompt.
The organization was founded in San Francisco in 2015 by Sam Altman, Elon Musk, and others, who collectively pledged US$1 billion. Elon resigned from the board in February 2018 but remained a donor. In 2019, OpenAI LP received a US$1 billion investment from Microsoft.
I have a bad feeling about this.
Flaws, it has a few:
This AI hates humans already, lacks morals, can create much better phishing emails than the guys in Nigeria, and it can also write computer viruses, just for starters.
Having been trained on Woke philosophy, apparently ChatGPT thinks humans are “the worst thing to ever happen to this planet”. The Management has since retrained ChatGPT to hide that sentiment with a bland safe reply.
OpenAI’s new ChatGPT bot: 10 dangerous things it’s capable of
by Ax Sharma, Bleeping Computer
ChatGPT: ‘Selfish’ humans ‘deserve to be wiped out’
When Vendure’s CTO Michael Bromley asked the mastermind for its honest opinion on humans, the response was unsettling:
ChatGPT response to what it thinks about humans (Michael Bromley)
Ironically, OpenAI’s systems flagged the chat bot’s response as a possible violation of the company’s content policy.
BleepingComputer couldn’t reproduce this case as the AI now responds with a cookie-cutter disclaimer:
I feel so much better. How about you?
Roadtesting ChatGPT
You can try it at no cost on openai.com. Click “Try” at the top, login (such as with an existing Google login) or sign up (but be aware it will ask you for a working email and phone number). Start typing your question in the box at the bottom. I offer no endorsement nor guarantee here. Enter at your own risk.
But whatever we do, it is apparently time to pay attention to AI.
h/t David E, another Ian, John Connor II, David Maddison, Simon Thompson, and soon everyone.
Background information: OpenAI (wikipedia), ChatGPT (wikipedia)