Artificial intelligence HLMI

AI at a human level: the difficulty of predicting

Results from a poll of AI experts show that on average there is a 50% chance high-level machine intelligence (HLMI) will be reached within AI by 2060.

Taking predictions at face value is often very dangerous and some sort of caution should be attached to them. As the well-worn saying goes, “Predictions are dangerous, especially about the future.” While a recent survey of 352 machine learning experts by Yale University and Oxford’s Future of Humanity’s Institute highlights their optimism about artificial intelligence’s development, it also indicates the general uncertainty about how rapidly AI will reach maturity.

The findings from the study show that on average there is a 50% chance of high-level machine intelligence (HLMI) – artificial intelligence that can carry out most human professions at least as well as a typical human – by 2060 and a 50% chance that they will take over all human jobs by 2136.

Researchers also predicted that AI will outperform humans in a number of activities within the next ten years including:

  • Translating languages (2024)
  • Writing high school essays (2026)
  • Driving a truck (2027)

Before examining the results of the report, it is worth bearing in mind that AI was not expected to be better than humans at the board game Go until 2027 but the world’s number one player was defeated this year. Indeed, twelve years ago two MIT researchers were adamant that AI would never replace human drivers. A lot can change in such a short space of time.

Given the thoughts of the people at the very top of the technological revolution such as Elon Musk, it is arguable that HLMI could arrive far sooner than we think. This should be treated in a positive light.

“I may be surprised, but I don’t think I will be”

The survey itself reveals the general uncertainty surrounding the issue of when AI will surpass human performance. First of all, the findings above are slightly misleading as the two questions are essentially the same. As the author makes clear, the two questions were included to explore if there was a framing effect within the experts’ predictions. The disparity in the median between the two answers certainly illustrates that there is a huge difference in opinion about the expected maturity date of AI.

It is also interesting to note that there are geographic differences when it comes to predicting HLMI. While experts from North America gave an aggregate prediction of 74 years, researchers from Asia believed it will happen in 30 years.

When considering other surveys on the predicted timing of human-level AI, the uncertainty becomes even more apparent. The median estimate for when there will be a 50% chance of human-level AI range from 2035 to 2050 according to seven surveys. In addition, the Bainbridge survey gave a median prediction of 2085 for HLMI. In short, no one really knows when to expect HLMI but there is an agreement that it will happen.

Why we should be optimistic about AI

Researchers, on the whole, were cautiously optimistic about the long-term benefits of AI. Of those surveyed, 85% gave answers ranging from neutral to extremely good when predicting AI’s final impact. Furthermore, 47% of respondents believe in further prioritizing safety research while 41% say it should continue at the same level. It is clear that those at the top of the field are wary of any potential pitfalls as AI continues to develop while retaining a certain level of confidence about its positive impact. This is a healthy way to view it.

An often-cited concern over AI’s inevitable rise is its impact on the job market.  There is a relevant school of thought that believes AI will create more jobs and improve education – as reflected on here. In fact, we have already seen tangible examples of how AI will benefit society.

How businesses can benefit:

For businesses to adapt in a world where disruption and change are the new norm they could follow the lead of Amazon which has used automation to scale its warehousing and shipping in the holiday season.

Its workforce expands by around 40% in the holiday season with the 120,000 new workers trained quickly thanks to AI tools such as automated training screens and robotic pallets. This means new hires are trained in as little two days compared with six weeks before.

Rather than being a job killer, AI can be a job creator if it is deployed to retrain displaced workers to this effect. Amazon ended up retaining 14% of their seasonal hires on a permanent basis as AI processed more total orders and e-commerce led to further job creation – Amazon built 26 new warehouses in 2016 alone.

As Deloitte outlines, businesses should also approach the concept of a job holistically and consider what value a human adds to each task. A bank teller can now advise and sell to clients instead of simply performing transactions. As Google’s Peter Norvig has argued, truck drivers will still be needed to load, unload and complete the paperwork. AI will remove repetitive tasks, not the jobs themselves.

To do this, journey maps can be developed to document the work taking place within each role in order to understand where tasks can be outsourced and where human empathy can be utilized further to make the role more effective.

The future of work is fast approaching and those who adapt quickly to the new landscape will thrive and remain relevant in an increasingly competitive market as we move closer to a world where colleagues are both human and robotic.

Inbenta utilizes its patented natural language processing and +11 years of research & development to create interactive chatbots with an industry leading +90% self-service rate.

Companies around the world including Ticketmaster UK utilize the InbentaBot to maintain a personal service for their customers while reducing support tickets.

Interested in finding out more? Our team of experts is at your service to design a custom proposal for you.

Let’s get in touch