Does the hype around generative AI reflect its technology progress?


ChatGPT’s release to the public in November 2022 started a wave of global interest and funding, but has the hype outrun AI development?

By Alice Nunwick

Since ChatGPT’s release in November 2022, generative AI has entered public discourse across the world. According to GlobalData, over a million social media posts about artificial intelligence (AI) have been made across Twitter and Reddit in the last year.

Generative AI has been rolled out to provide customer relationship management solutions, software development and even storytelling. But some sceptics cannot ignore that the timing of generative AI’s hype is fortuitous for the tech industry.

Managing director at TS Lombard, Dario Perkins, explains in a recent webinar that Big Tech was “the part of the stock market that suffered the largest declines as central banks started aggressively raising interest rates”.

Perkins also attributes the “drying up” of tech investment to the closure of several banks with close ties to the tech sector, such as Silicon Valley Bank, and describes ChatGPT’s public release as “very clever marketing”.

With the metaverse winter closing in and quantum computing funding slowing down due to an emerging lack of practical uses, generative AI instead began experiencing a funding frenzy whilst funding in AI overall dropped. On 13 June, French company Mistral AI broke the European record for a seed round funding, receiving a total of $260m in four weeks.

AI has also boosted hiring for tech companies who Perkins notes were seeing “fairly big job losses in contrast to other sectors- in contrast to other sectors who were still looking to add jobs”. GlobalData research consolidates that the tech sector’s hiring in generative AI increased approximately 600% from March 2023 to June 2023.

Alongside the rush of news and social media coverage, and the increase in hiring, a recent working paper by MIT proposes that the use of ChatGPT was able raise efficiency by almost 40% and greatly improve the quality of work by lower achieving students.

GlobalData’s tech sentiment polls indicate that AI was perceived as the most disruptive technology in the last quarter of 2022. Despite this numerous companies such as Apple, JP Morgan, Deutsche Bank and Verizon have all banned the use of the generative AI in 2023.

Does generative AI’s lack of understanding hinder its progression?

As search engines rolled out to optimize internet use, it is no surprise that people have asked whether the ability to just ‘google’ a question has been making our brains slower and less retentive of information. This same question is now being asked of AI.

In response to these worries, Global Data’s Research Director Josep Bori states that while current generative AI can be an “extremely useful and powerful tool”, they are “probably not for totally unsupervised factual advice anytime soon”.

Another challenge for generative AI are large language model (LLM) hallucinations. Speaking about this problem being fixed soon, Bori remains pessimistic, citing that LLMs are not able to fact check themselves for answers that are “falsehoods or plainly absurd”. Both Bori and Perkins emphasize that generative AI tools have no understanding of the world or the data they are trained on. “Philosophically speaking,” states Perkins, “it’s bullshit”.

Although there is a steady rise in the use of generated content in online news sites, this has detrimentally increased the possibility of misinformation, and the risk of legal action should a company be wrongly mentioned in these hallucinations.

When speaking on whether AI will ever be able to verify its own answers, Bori asks us to consider the method of scientists.

“It starts with formulating a plausible hypothesis,” begins Bori, “then real experiments either confirm or invalidate those hypotheses, and the cycle goes on.” But since LLMs cannot test their knowledge, a route to creating an AI that can corroborate its answers remains unclear.

Bori remains skeptical of the possibility of LLMs becoming more than “limited memory machines”. Without a proper roadmap towards generative AI that produces plausible answers every time, generative AI faces becoming another investment trend of the past.