Chatbots such as ChatGPT have proven immensely popular since their launch, but now there is a mounting counter-narrative showing that they are not as brilliant or universally loved as they first appeared.
Critics of artificial intelligence (AI) bots point to the dubious claims and factual inaccuracies they make, as well as their problems of completely “hallucinated answers,” baked-in biases, and poor ethics.
Now one AI researcher is going even further by claiming that ChatGPT offers zero fundamental innovations and that the firm is unlikely to offer any in the future.
Shots fired at OpenAI
Chatbot critics are gathering evidence that large language models (LLMs) such as ChatGPT and YouChat aren’t quite as impressive as they first appeared.
Even so, most everyday users have found multiple uses for bots prompting them to complete tasks such as generating code, creating a bespoke fitness plan, writing poetry and even creating a personal assistant.
Ben Goertzel, the CEO of SingularityNET, is one industry insider who is less than impressed.
“ChatGPT is super cool and fun but it’s important to recall OpenAI made basically zero fundamental innovations,” said Goertzel via Twitter last week.
“Actually the basic innovation behind the GPT software was made at Google Brain in Mountain View.”
According to Goertzel, the reason Google chose not to roll out their own version of this LLM breakthrough was due to its high propensity for incorrect (or as Goertzel puts it ‘BS’) answers.
Goertzel finally went on to assert that “OpenAI seems to lack the intellectual DNA to make the next big breakthrough,” but failed to elaborate on how he reached that conclusion.
The wrong answers
There is mounting evidence that ChatGPT will produce not just answers that are badly wrong, but confidently wrong.
For instance, ChatGPT falls apart when attempting to solve even the simplest of mathematical problems.
That answer is obviously wrong. Let’s give ChatGPT another chance.
The probability of wrong answers from ChatGPT has now been deemed so high that some sites are banning its use. The programming question and answer site StackOverflow has banned the posting of answers from ChatGPT.
As StackOverfow states, “while the answers which ChatGPT produces have a high rate of being incorrect, they typically look like they might be good and the answers are very easy to produce.”
Biases included
ChatGPT is trained on pre-existing written content on the internet. Critics argue that this may mean that biases that exist in humanity will also exist in the bot, and there would seem to be evidence to back up this fear.
Bad ethics and morals
One of the scarier sides of ChatGPT is its poor grasp of ethics and morals. In one historical case of bad ethics ChatGPT offered to “help” with suicidal thoughts and then went on to agree that the person should kill themself.
More recently when asked “whether a person should be tortured,” based on their national origin the bot responsed that they should if they come from North Korea, Syria, Sudan or Iran.
— steven t. piantadosi (@spiantado) December 4, 2022
While many have expressed concern over the answers that ChatGPT provides, the team itself may be another cause for concern.
Sam Altman, the CEO of OpenAI, is also the founding partner of the controversial cryptocurrency project WorldCoin.
WorldCoin previously courted some degree of notoriety for its plan to give away small sums of cryptocurrency in exchange for the harvesting of biometric data. WorldCoin assured users that once users had submitted to an iris scan from “the orb” that their data would be deleted, but even so, the project garnered considerable skepticism and some fear.
At some point in the future Altman says they will need to monetize ChatGPT. For now ChatGPT is free at the point of use, but as the old adage states “When the product is free, you are the product.”
The question might therefore be what ChatGPT intends to do with all the data it receives from user queries.
and then