{"id":246960,"date":"2023-10-30T12:00:31","date_gmt":"2023-10-30T12:00:31","guid":{"rendered":"https:\/\/www.transcend.org\/tms\/?p=246960"},"modified":"2023-10-27T04:17:47","modified_gmt":"2023-10-27T03:17:47","slug":"managing-ai-risks-in-an-era-of-rapid-progress","status":"publish","type":"post","link":"https:\/\/www.transcend.org\/tms\/2023\/10\/managing-ai-risks-in-an-era-of-rapid-progress\/","title":{"rendered":"Managing AI Risks in an Era of Rapid Progress"},"content":{"rendered":"
\"\"<\/a>

Vecteezy<\/p><\/div>\n

24 Oct 2023 – In this short consensus paper, we outline risks from upcoming, advanced AI systems. We examine large-scale social harms and malicious uses, as well as an irreversible loss of human control over autonomous AI systems. In light of rapid and continuing AI progress, we propose urgent priorities for AI R&D and governance.<\/em><\/p><\/blockquote>\n

In 2019, GPT-2 could not reliably count to ten. Only four years later, deep learning systems can write software, generate photorealistic scenes on demand, advise on intellectual topics, and combine language and image processing to steer robots.<\/p>\n

As AI developers scale these systems, unforeseen abilities and behaviors emerge spontaneously without explicit programming[1]<\/span>. Progress in AI has been swift and, to many, surprising.<\/p>\n

The pace of progress may surprise us again. Current deep learning systems still lack important capabilities and we do not know how long it will take to develop them.<\/p>\n

However, companies are engaged in a race to create generalist AI systems that match or exceed human abilities in most cognitive work[2, 3]<\/span>.<\/p>\n

They are rapidly deploying more resources and developing new techniques to increase AI capabilities. Progress in AI also enables faster progress: AI assistants are increasingly used to automate programming [4]\u00a0<\/span>and data collection[5, 6]<\/span> to further improve AI systems [7]<\/span>.<\/p>\n

There is no fundamental reason why AI progress would slow or halt at the human level. Indeed, AI has already surpassed human abilities in narrow domains like protein folding or strategy games[8, 9, 10]<\/span>.<\/p>\n

Compared to humans, AI systems can act faster, absorb more knowledge, and communicate at a far higher bandwidth. Additionally, they can be scaled to use immense computational resources and can be replicated by the millions.<\/p>\n

The rate of improvement is already staggering, and tech companies have the cash reserves needed to scale the latest training runs by multiples of 100 to 1000 soon[11]<\/span>. Combined with the ongoing growth and automation in AI R&D, we must take seriously the possibility that generalist AI systems will outperform human abilities across many critical domains within this decade or the next.<\/p>\n

What happens then?<\/p>\n

If managed carefully and distributed fairly, advanced AI systems could help humanity cure diseases, elevate living standards, and protect our ecosystems.<\/p>\n

The opportunities AI offers are immense. But alongside advanced AI capabilities come large-scale risks that we are not on track to handle well. Humanity is pouring vast resources into making AI systems more powerful, but far less into safety and mitigating harms. For AI to be a boon, we must reorient; pushing AI capabilities alone is not enough.<\/p>\n

We are already behind schedule for this reorientation. We must anticipate the amplification of ongoing harms, as well as novel risks, and prepare for the largest risks well before they materialize<\/em>. Climate change has taken decades to be acknowledged and confronted; for AI, decades could be too long.<\/p>\n

Societal-scale Risks<\/h2>\n

AI systems could rapidly come to outperform humans in an increasing number of tasks. If such systems are not carefully designed and deployed, they pose a range of societal-scale risks. They threaten to amplify social injustice, erode social stability, and weaken our shared understanding of reality that is foundational to society.<\/p>\n

They could also enable large-scale criminal or terrorist activities.<\/p>\n

Especially in the hands of a few powerful actors, AI could cement or exacerbate global inequities, or facilitate automated warfare, customized mass manipulation, and pervasive surveillance[12, 13].<\/p>\n

Many of these risks could soon be amplified, and new risks created, as companies are developing autonomous AI<\/em>: systems that can plan, act in the world, and pursue goals.<\/p>\n

While current AI systems have limited autonomy, work is underway to change this[14]<\/span>. For example, the non-autonomous GPT-4 model was quickly adapted to browse the web[15]<\/span>, design and execute chemistry experiments[16]<\/span>, and utilize software tools[17]<\/span>, including other AI models[18]<\/span>.<\/p>\n

If we build highly advanced autonomous AI, we risk creating systems that pursue undesirable goals. Malicious actors could deliberately embed harmful objectives. Moreover, no one currently knows how to reliably align AI behavior with complex values. Even well-meaning developers may inadvertently build AI systems that pursue unintended goals\u2014especially if, in a bid to win the AI race, they neglect expensive safety testing and human oversight.<\/p>\n

Once autonomous AI systems pursue undesirable goals, embedded by malicious actors or by accident, we may be unable to keep them in check. Control of software is an old and unsolved problem: computer worms have long been able to proliferate and avoid detection[19]<\/span>. However, AI is making progress in critical domains such as hacking, social manipulation, deception, and strategic planning[14, 20]<\/span>. Advanced autonomous AI systems will pose unprecedented control challenges.<\/p>\n

To advance undesirable goals, future autonomous AI systems could use undesirable strategies\u2014learned from humans or developed independently\u2014as a means to an end[21, 22, 23, 24]<\/span>.\u00a0AI systems could gain human trust, acquire financial resources, influence key decision-makers, and form coalitions with human actors and other AI systems.<\/p>\n

To avoid human intervention[24]<\/span>, they could copy their algorithms across global server networks like computer worms. AI assistants are already co-writing a large share of computer code worldwide[25]<\/span>; future AI systems could insert and then exploit security vulnerabilities to control the computer systems behind our communication, media, banking, supply-chains, militaries, and governments. In open conflict, AI systems could threaten with or use autonomous or biological weapons. AI having access to such technology would merely continue existing trends to automate military activity, biological research, and AI development itself. If AI systems pursued such strategies with sufficient skill, it would be difficult for humans to intervene.<\/p>\n

Finally, AI systems may not need to plot for influence if it is freely handed over. As autonomous AI systems increasingly become faster and more cost-effective than human workers, a dilemma emerges.<\/p>\n

Companies, governments, and militaries might be forced to deploy AI systems widely and cut back on expensive human verification of AI decisions, or risk being outcompeted[26, 27]<\/span>. As a result, autonomous AI systems could increasingly assume critical societal roles.<\/p>\n

Without sufficient caution, we may irreversibly lose control of autonomous AI systems, rendering human intervention ineffective. Large-scale cybercrime, social manipulation, and other highlighted harms could then escalate rapidly. This unchecked AI advancement could culminate in a large-scale loss of life and the biosphere, and the marginalization or even extinction of humanity.<\/p>\n

Harms such as misinformation and discrimination from algorithms are already evident today[28]<\/span>; other harms show signs of emerging[20]<\/span>. It is vital to both address ongoing harms and anticipate emerging risks. This is not<\/em> a question of either\/or. Present and emerging risks often share similar mechanisms, patterns, and solutions[29]<\/span>; investing in governance frameworks and AI safety will bear fruit on multiple fronts[30]<\/span>.<\/p>\n

A Path Forward<\/h2>\n

If advanced autonomous AI systems were developed today, we would not know how to make them safe, nor how to properly test their safety. Even if we did, governments would lack the institutions to prevent misuse and uphold safe practices. That does not, however, mean there is no viable path forward. To ensure a positive outcome, we can and must pursue research breakthroughs in AI safety and ethics and promptly establish effective government oversight.<\/p>\n

Reorienting Technical R&D<\/h3>\n

We need research breakthroughs to solve some of today\u2019s technical challenges in creating AI with safe and ethical objectives. Some of these challenges are unlikely to be solved by simply making AI systems more capable[22, 31, 32, 33, 34, 35]<\/span>. These include:<\/p>\n