Article: Get over it, your AI is not synonymous with accuracy

Technology

Get over it, your AI is not synonymous with accuracy

Fast, precise, and logical are some of the descriptive terms attached with AI. So much so that it has conveniently replaced the human mind when accuracy is desired. But with ChatGPT, Bing, and Bard slowly snuggling into our lives, we cannot go long without addressing the contradiction at the heart of AI - its unreliability.
Get over it, your AI is not synonymous with accuracy
 

A complete dependence on AI for facts and information can be delusional.

 

Calculators, GPS systems and browsers flooded with answers for your one question, the dependence on AI to seek accurate answers has become almost a universal truth. It is this daily psychology of users that requires challenging when AI systems advance and their participation in our lives don't remain as simple anymore.

The first two weeks of February 2023 was a bench-marking time for the tech world as leading tech companies, Google and Microsoft, introduced generative AI technology to their search engines. It is not far-fetched to imagine ChatGPT, Bard and Bing becoming an inseparable part of the everyday. The pace at which it is making space in the lives of users is terrifyingly fast. Twitter is full of conversations related to this new world of wonder. Job requirements in various platforms have already started to appear for people well-versed in this technology.

Bill Gates, in an interview, said, “This (technology like ChatGPT) will change our world…the progress over the next couple of years to make these things even better will be profound.”

The question this seemingly utopian situation triggers is whether this wonder brings along woe. Is this new segment of artificial intelligence completely reliable as we are on the rest almost unconsciously?

Alphabet, Google’s parent company, has already lost $100 billion in market value on Wednesday due to a factual error caused by its AI technology in a promotional tweet. 

In December of 2022, Elon Musk, posted a tweet saying, “ChatGPT is scary good. We are not far from dangerously strong AI.”

Elon Musk's tweet about ChatGPT

Then, how scary? How dangerous?

The father of the internet and ex-vice president of Google, Vint Cerf said, “There’s an ethical issue here that I hope some of you will consider.” Adding on to the conversation he claimed AI to be “really cool, even though it doesn’t work quite right all the time.”

In a world where accuracy, correctness and unbiasedness is very often equated with AI-generated solutions or answers, it is important to understand that the situation is more complex than ever. The technology through which these intelligences operate (large language models or LLMs) is said to “generate bullshit.”

It tries to produce an answer that is continuously reasonable after gathering information from other web pages. Hence, to summarise, ChatGPT, Bing and Bard are not functioning in a vacuum but are directed and influenced by human guidance. 

Several tweets have accused it of operating from a left-leaning ideology while the others have claimed to be running under a sexist and racist command. For instance, when asked to write a poem, it delivered on Biden but not Trump. Accusations like these are increasing on the internet making it important to point out that a complete dependence on AI for facts and information can be delusional.

Although the above argument refers to a more interpretive point of view, AI technologies haven't been quite successful in delivering the right, fact-based, objective information either. With this recognition comes the understanding of a conscious distance that needs to be maintained when consuming information from these platforms. 

Anubha Shrivastava Sahai, President of India Wide Parents Association and a lawyer, shares that ChatGPT incorrectly stated that she has been awarded with Nari Shakti Purashkar in the year 2020.

Anubha Sahai tweets about ChatGPT's error

Upol Ehsan accused ChatGPT of wrongly crediting the paper that he wrote. He adds an argument further saying, “The point isn’t to poke fun at ChatGPT or other LLMs. Of course they hallucinate. The point is how plausible sounding these hallucinations are and what it means for us being able to separate fact from fiction.”

Upol Ehsan tweets about ChatGPT wrongly crediting his paper.

Even after being incorrect at several times and levels, a lot of instances have occurred where the AI not only rejects being corrected but even blames the user for the error. The technology's lack of flexibility in learning and its static idea about the world gaslights the user and will only end up in disappointment on the user’s part. 

Jon Uleis tweets about ChatGPT getting the year wrong

 

A Twitter user shares a screenshot of ChatGPT giving misinformation.

Hence, it's crucial to be hyper aware about the answers one is receiving while in communication with these bots. Sometimes the misinformation will be on the face and too obvious to miss. On other times, it will be well-organised and convincing enough to lure one into believing it as true.

At the end of the day, generative AI technology absolutely requires the intervention of the user when consuming the information and marking both the truth and propaganda, if any, in it.

Read full story

Topics: Technology

Did you find this story helpful?

Author


QUICK POLL

How do you envision AI transforming your work?