‘The science is here’: UN chief welcomes first global AI assessment

“The science is here,” UN Secretary General Antonio Guterres said at the launch of the report. “We can no longer say that we don’t know. What we do with it now is up to all of us.”

The more progress AI achieves without common rules, the less opinion it will generate among governments and society, said the UN Secretary General, adding “My message to the government is simple: don’t wait.”

Aiming to build shared understanding and evidence at this critical time, the Initial Report of the Independent International Scientific Panel on AI: An evidence-based assessment of the opportunities, risks and impacts of AI written by the first fully independent global scientific body dedicated to assessing real impacts on the economy and society.

Read the full report Here.

Why it matters

Globally, more than a billion people now use conversational AI every week, while governments make critical decisions amid great uncertainty due to rapidly changing and often conflicting sources of evidence and perspectives that do not reflect local realities.

“Used well, AI could be the most powerful engine for development, accelerating global progress on everything from health and hunger to learning and climate,” said the UN Secretary General, “but the panel also clearly understands the devastating impact that artificial intelligence can have.”

It’s true, as AI’s capabilities grow, so do the challenges it faces – which is the main challenge this panel aims to address.

Read our AI explainer here.

A better world or a major disaster?

Comprising 40 leading scientists and experts from each region, the panel outlines AI trends and warns that current safeguards are not keeping pace, said the panel’s co-chair, Yoshua Bengio.

“AI’s capabilities surpass scientific understanding and governments’ ability to adaptBengio said. “With increasing evidence of deceptive AI behavior, current science cannot guarantee that as capabilities improve, AI will not cause major harm, whether by itself or by malicious users.”

To act effectively, he said, global policymakers must understand these systems, and the panel provided just that: a strong common scientific foundation “to guide our collective path forward.”

Main findings

Detecting breast cancer earlier, accelerating vaccine development, and improving healthcare are just some of AI’s breakthrough achievements, but limitations and challenges remain, including:

• AI adoption has increased widely, but not evenly across countries and sectors
• Access and use vary widely, with adoption in southern countries lagging far behind that in northern countries
• Significant differences in computing infrastructure and models exist among developed countries, reflecting existing disparities

Moreover, development is highly concentrated, with recent estimates showing that the United States accounts for 75 percent of the computing power among the world’s 500 largest AI supercomputers, while China accounts for 15 percent, and that companies in both countries develop nearly all of the leading general-purpose models.

Understand the risks

Understanding and managing AI risks is critical, the report said, and panel co-chair Maria Ressa added that the risks to society, security and the human species are already “too high”.

“This technology is transformative, but if the world continues to follow this trend, humanity will fail to realize its promised benefits,” he warned.

Here are some panel warnings:

• There is no scientific guarantee that an AI agent system will not violate instructions, and evidence is mounting of cases that have already occurred
• AI agent systems will soon complete tasks that currently take human programmers days or weeks, but their implementation raises pressing questions about the labor market, cybersecurity, and controllability of future AI systems.
• Sycophantic AI behavior, i.e. responses that reinforce a user’s beliefs, regardless of their accuracy, has been linked to several serious mental health incidents, including documented deaths.
• Criminals and malicious actors have been documented using AI systems to aid cyberattacks
• Advanced technical capabilities enable budding private actors to use AI in nefarious ways in applications such as fraud and disinformation
• Reliable methods for maintaining control over highly autonomous AI systems are still lacking

The report states that many of these adverse impacts occur disproportionately on already disadvantaged groups of society.

Unlock benefits, mitigate risks

Minimizing AI risks and leveraging these technological tools requires good governance.

There are concrete next steps to close existing gaps, but each requires continued investment in Member States’ capacities to shape, evaluate and implement AI.

Realizing these opportunities safely requires specific investments and policies to incentivize equitable access and reward innovation while preventing exploitation of vulnerable groups.

‘AI will not close the divide by itself’

The report found that dozens of different governance instruments that seek to incorporate ethics and human rights into AI systems are already in use in various jurisdictions, but they are fragmented, concentrated in a few companies, and rarely measure real-world effectiveness.

Amandeep Gill, Deputy Secretary-General and Special Envoy for Digital and Emerging Technologies, said the new report conveys scientific language that can be used by future decision makers.

“AI will not close the gap by itself,” he said.

The benefits will be felt when the institutions, skills and data are in place, and if they are not there, the same technology can displace workers, widen gaps and make people dependent on systems built without them knowing, he explained.

“That reality is now on record, independently verified, and impossible to set aside.”

The report’s findings will be presented to governments at the first-ever UN Global Dialogue on AI Governance, held in Geneva on July 6 and 7.

Watch the report launch and press conference here: