Amid the mainstream boom, AI skeptics say the technology is overhyped.

Artificial intelligence tools like ChatGPT will “clearly become mainstream” by 2023, according to a new survey of more than 1,500 professionals working in technology and related fields. However, despite rapid adoption, most respondents still believe AI is overrated. Less than a quarter rated it ‘fairly appraised’.

A survey conducted by Retool found widespread enthusiasm tempered by skepticism. Founded in June 2017 and part of the Y Combinator accelerator program, Retool provides a platform for building internal business software using a visual development interface.

“By 2023, AI, especially generative AI, will be firmly established as mainstream,” the report said. “It seems like everyone has used large-scale language models (LLMs). With over 100 million active weekly users, ChatGPT has become a virtual household name.”

Teachers, talk show hosts and even grandparents are now using AI chatbots for more than just laughs, the survey shows. But real-world applications only “scratch the surface” of AI’s potential.

More than half of respondents (51.6%) still rate AI as overvalued, while only 23.4% rate it as adequate. Top management and executives expressed the most favorable views. The software engineers closest to the implementation were even more skeptical.

AI Transforms Jobs and Industries

Respondents agreed that AI will substantially transform industries and jobs within five years. On an impact scale of 0 to 10, the average score exceeded 7. People in operational roles expected the highest level of change, followed by product managers and engineers.

Expectations were focused on four key areas: increased efficiency, reduced tedious work, new learning technologies, and updated design processes. But respondents also want assurance that AI will be developed safely.

“They also asked the big, important question of investing in AI governance and ethics to ensure the future of the technology is positive,” the report urges.

Dependency on Stack Overflow Drop

Many engineers are now relying less on the popular Stack Overflow programming site for coding help. Nearly 60% of survey respondents will use Stack Overflow less after 2022. Of those, one in 10 stopped using it completely.

What is the biggest reason for the decline? GitHub Copilot and ChatGPT. Together they accounted for nearly 94% of the responses.

Launched in 2021, GitHub Copilot provides contextual code suggestions as programmers type. The AI ​​tool works as an automated twin programmer.

Do you need AI technology now?

When using ChatGPT and GitHub Copilot to rate the likelihood of hiring a candidate who is proficient at writing code, respondents scored an average of 6.7 on a 10-point scale. Small startups showed the greatest openness, while midsize companies were more cautious.

However, nearly 30% still said they were neutral about AI coding technology. Respondents rarely saw this as a negative thing.

“The ability to write code using AI had some positive impact for the majority (63.2%), but a significant number of respondents (27.1%) remained neutral,” the report said.

What’s behind enterprises’ hunger for AI?

Survey respondents believe cost savings and buzz are what drive most companies to pursue AI. The most cited motivations include cost savings (38.7%), pandemic (35.2%), customer demands (32.9%), and competitive pressures (31.1%).

In contrast, only 25.5% cited direct revenue impact as their primary driver. The report suggests that, at least for now, companies may view AI as a way to reduce costs rather than increase revenue.

AI is useful but inaccurate

Companies actively using AI tools rate them as generally useful, although not perfect. Issues with accuracy, data security and apparent hallucinations top the list of complaints.

“For whatever purpose, today’s AI tools are unlikely to be perfect,” the report wisely observes. “But understanding the pros and cons of any technology can help you position yourself to use that technology well.”

Nearly 40% of survey respondents cited “model output accuracy” as a major challenge in AI app development, with “hallucinations” coming in third at 28.1%. Data security falls somewhere in between at 33.4%.

Will OpenAI’s dominance last?

For enterprises implementing AI, OpenAI’s natural language services have higher adoption rates than other providers. Different types of ChatGPTs constitute the most used models by respondents.

But sticking to off-the-shelf products still outweighs customization. Companies that use self-hosted open source algorithms rather than using packaged services remain an exception for now.

behind the numbers

Insights from the Retool report come from a public survey of 1,578 respondents conducted in August. Technology workers made up 39% of participants, followed by consulting and professional services at 12% and financial services at 10%.

More than a third classified themselves as engineers, more than a fifth in an operations role, and 12% as product managers. 17% held top management positions such as CEO, 28% held mid-senior management positions, and 20% held entry-level positions.

Companies with 1 to 99 employees accounted for 60% of the total, medium-sized companies with 100 to 999 workers accounted for 26%, and large companies with 1,000 or more workers accounted for the remaining 14%.

Editor’s note: This story was drafted using Decrypt AI from sources referenced in the text. fact confirmed Written by Ryan Ozawa.