Crafting safe Generative AI systems

Source– The post is based on the article “Crafting safe Generative AI systems” published in “The Hindu” on 21st August 2023.

Syllabus: GS3- Awareness in the field of IT and computers

News– The article explains the potential of generative AI and risks related to it. It also explains the importance of identity assurance framework to deal with the risk associated with generative AI.

What is the potential of generative AI?

Generative AI has the potential to unleash significant technological and societal transformations. Utilization of Large Language Models alone is projected to contribute between $2.6 trillion and $4.4 trillion annually to the global economy.

Example is Jugalbandi Chatbo. It aims to function as a universal translator in rural India. It receives queries in local languages, and then presents these responses to users in their native tongue.

This has the capacity to democratize access to information and enhance the economic well-being of millions of individuals.

What are risks related to generative AI?

AI-powered tools enable malicious actors to craft synthetic entities. These cannot be distinguished from humans online, using speech, text, and video formats.

These bad actors can manipulate identities. They can launch a range of harmful actions such as spreading misinformation, disinformation, engaging in security breaches, perpetrating fraud, propagating hate speech, and public shaming.

In the United States, an AI-generated image depicting the Pentagon ablaze caused turmoil in equity markets.

Fake users on platforms like Twitter and Instagram, disseminating strong political opinions, have exacerbated online political polarization.

AI-generated voices have been cloned and utilized to bypass authentication measures for banking customers.

An individual in Belgium was reportedly driven to suicide after interactions with a Large Language Model. Additionally, recent elections in Turkey were marred by AI-generated deepfakes.

The risk of malicious actors exploiting Generative AI for purposes of misinformation and influencing election outcomes is steadily on the rise.

What is the potential solution to risks related to generative AI?

Regulation alone is insufficient. A broader perspective must be taken into account to enhance Internet safety and reliability.

There is a need for an identity assurance framework. This framework ensures trust among interacting parties by verifying the authenticity of entities involved.

What are some facts about the identity assurance framework?

The core principles of this framework involve its inclusion of various emerging credential types worldwide. It does not have exclusive reliance on any single technology or standard.

Digital wallets are important. They offer selective disclosure and protect users from governmental or corporate surveillance. This identity assurance framework would encompass humans, bots, and businesses.

At present, more than 50 countries are engaged in initiatives to develop or issue digital identity credentials. India, with Aadhaar, holds a leading position in implementing online identity assurance safeguards.

The European Union is in the process of establishing a new identity standard that will also support online identity assurance.

The concept of identity assurance is intertwined with the issue of information integrity. Information integrity guarantees that the accessed content is genuine and originates from the purported source.

This credibility is founded upon three pillars. Firstly, source validation ensures the verification of the information’s origin, publisher, or individual. Secondly, content integrity guarantees that the information remains unaltered.

Lastly, information validity can be achieved through automated fact-checking and crowd-sourced reviews.

Identity assurance faces dilemmas like balancing privacy and surveillance, civil liberties and security, and anonymity and accountability.

Information integrity raises questions about censorship and the timeless query of ‘who defines the truth?’