Microsoft says deepfakes are AI’s biggest problem

Microsoft President Brad Smith expressed concern about the broad spread of realistic-looking but false content. The Guardian reports this.

\n\n

During a Washington address on AI regulation, Smith urged steps to detect deepfakes.

\n\n

\”We will have to address the problems associated with deepfakes. We will have to pay particular attention to the fact that most foreign operations of cyber influence are already carried out by the Russian government, the Chinese, and the Iranians,\” he said.

\n\n

Smith believes that steps should be taken now to guard against AI-enabled manipulation of legitimate content and deceive people. He also called for licensing the most important forms of artificial intelligence with safety commitments.

\n\n

\”We will need a new generation of export controls […] to ensure that our models do not steal or misuse in violation of requirements,\” he said.

\n\n

Alignment with OpenAI

\n\n

Last week OpenAI CEO Sam Altman testified before the U.S. Congress, where he called for AI regulation. The company’s CEO also spoke of possible licensing of the most important technologies to ensure developers meet certain criteria.

\n\n

After that Altman went to Europe, where he met with leaders of the United Kingdom, France, Germany and other countries. There he stated that OpenAI would have to exit the EU market if the company cannot meet the AI Act’s ‘stringent provisions’.

\n\n

Many AI experts and thought leaders support the move to regulate the technology. In March they signed an open letter calling for a pause on experiments with large language models.

\n\n

However the initiative faced criticism. Microsoft co-founder Bill Gates believes that a pause will not help to tame AI.

\n\n

According to the billionaire, slowing AI globally is becoming increasingly difficult. Instead, the world should focus on using the technology for the benefit of society.

\n\n

Altman backed only some of the theses of the open letter.

\n\n

Microsoft’s efforts to combat deepfakes

\n\n

At Build 2023, Microsoft unveiled several services to combat deepfakes. In the coming months, the company will roll out the C2PA specification to verify images and videos for use with Bing Image Creator or Designer.

\n\n\n\n

Microsoft also introduced the Azure AI Content Safety content moderation tool. The service is trained to detect toxic images and texts generated by both humans and AI.

\n\n

The models operate in multiple languages. They assign a content risk level and alert the moderator to take action as needed.

\n\n

Earlier in May, former Google CEO Eric Schmidt called AI an \”existential threat\”, from which many people could \”suffer or die\”.

\n\n

In the same month, billionaire investor Warren Buffett compared artificial intelligence to the creation of an atomic bomb and expressed concern about the rapid development of the technology.