1 hr ago
Microsoft President Brad Smith Backs Independent AI Safety Evaluators
Microsoft President Brad Smith said AI systems should be checked by independent safety experts.
He said safety should not be controlled by only one or two AI companies.
Microsoft wants software companies and governments to help monitor AI as well.
Smith said Microsoft and OpenAI use a safety board to help decide when models are ready to release.
He also supports an off-switch that lets people stop advanced AI systems.
Microsoft plans to use a large amount of computing power at a data center in the United Arab Emirates.
The company has invested $1.5 billion in G42, an Emirati technology firm.
Smith also said Microsoft may have to change its climate plans because AI data centers need more electricity.
Microsoft President Brad Smith said the company supports independent evaluators for artificial intelligence safety.
Smith argued that AI safety should involve multiple layers, including model companies, software firms and governments.
Microsoft has discussed model safety with OpenAI through a joint safety board and supports a human-controlled off-switch.
Microsoft plans to use 200 megawatts at an AI data center in the United Arab Emirates and has invested $1.5 billion in G42.
Smith said Microsoft may need to adjust its 2030 climate goals as AI data centers increase demand for natural-gas power.
- Who
- Microsoft President Brad Smith and Microsoft; the discussion also involved OpenAI, Anthropic and G42.
- What
- Smith backed independent AI safety evaluators, multilayered controls and human-accessible off-switches for advanced AI systems.
- Where
- The discussion coincided with the United Nations General Assembly in New York; the investment and data-center plans concern the Middle East and the United Arab Emirates.
- When
- Smith made the comments in an interview with Bloomberg Television on Wednesday, during the United Nations General Assembly.
- Why
- Smith said AI safety would not advance if responsibility were left to only one or two entities, and said additional computing demand may require changes to Microsoft’s climate plans.
Key facts
- AI safety position
- Microsoft supports independent evaluators and safety controls distributed across multiple layers.
- Microsoft-OpenAI safety work
- The companies use a safety board staffed by both organizations to help determine when models are safe to release.
- Human control
- Smith supports an off-switch that can keep advanced AI systems under human control.
- G42 investment
- Microsoft has invested $1.5 billion in Emirati technology firm G42.
- Planned computing capacity
- Microsoft still plans to use 200 megawatts at a planned AI data center in the United Arab Emirates.
- Climate goals
- Smith said Microsoft may need to adjust its goal of removing more carbon from the atmosphere than it emits by 2030.
- Energy challenge
- The article links the climate-goal difficulty partly to Microsoft’s use of natural-gas plants to power AI data centers.
Quotes
Brad Smith
Microsoft president discussing the roles of model developers and software companies in AI safety
“You really want safety to be built in multiple layers. At times, you may want the government to be able to have the on-off switch, but why would you put that switch in only one place when you can put it in three or four?”
livemint.com
“The model companies are obviously critical, but so are companies like ours that produce the layer of software that controls AI agents, that monitors what they do.”
livemint.com
Sources
Related news
Nadella Backs Deliberate AI Progress and Human Control
Microsoft AI Chief Rejects China Argument Against Regulation
Anthropic, Accenture Commit $2 Billion to AI Model Evaluation
Zuckerberg Says Trust and Alignment Will Shape AI Safety
Microsoft Drafts AI Code to Keep Systems Under Human Control
Microsoft AI Chief Calls OpenAI Safety Incident Serious
India-France-UAE AI Partnership Targets Jobs, Startups, Infrastructure
UAE Revises AI Data Center Plans After Iranian Attacks
Former Xiaomi India Head Manu Jain Alarmed by Delhi's Pollution
Musk's Attempt to Block OpenAI Deal in Middle East Fails
