What AI Can Do, Its Risks, and Mitigation Strategies from the International AI Safety Report   Recently updated !

This entry was posted in Newsletter on by

Welcome to your weekly AI Newsletter from AITechCircle!

I’m building and implementing AI solutions and sharing everything I learn along the way…

Check out the updates from this week! Please take a moment to share them with a friend or colleague who might benefit from these valuable insights!

Today at a Glance:

  • Key Actions from the International AI Safety Report
  • Generative AI Use cases repository
  • AI Weekly news and updates covering newly released LLMs
  • Courses and events to attend

Insights from the First International AI Safety Report

The first-ever International AI Safety Report is backed by 30 countries, including the OECD, UN, and EU, and has just been published.

This Report focuses on General-purpose AI, whose capabilities have rapidly advanced in the last few years.

It aims to inform policy discussions at the upcoming AI Action Summit in France, highlighting the rapid advancements in AI capabilities and the importance of managing emerging risks responsibly to leverage AI’s potential safely.

It answers three questions:

  1. What can general-purpose AI do?
  2. What are its risks?
  3. How can these risks be mitigated?

​

The capabilities of general-purpose AI models have increased tremendously over the years. The chart below shows the level of expertise.

​

AI performance vs human performance

The performance of AI models on various benchmarks has advanced rapidly between 1998 and 2024.

Risks

The report categorizes risks associated with general-purpose AI into three categories:

  1. Risks from malicious use

    • Harm to individuals through fake content
    • Attempts to manipulate public opinion
    • Cyberattacks / cyber offenses
    • Biological/chemical attacks
  2. Risks from malfunctions

    • Reliability issues & Hallucinations
    • Biased model outputs
    • Loss of control risk (Active loss of control or passive loss of control)
  3. Systemic risks

    • Labour market risks
    • Global AI R&D divide
    • Market concentration & single points of failure
    • Risks to the environment
    • Risks to privacy (training risks, use risks, intentional harm risks)
    • Risks of copyright infringement
    • Impact of open-weight general-purpose AI models on AI risks

Risks Management:

Evaluations must focus on broader risks from general-purpose AI, not just capabilities, to better inform risk management practices. Additionally, they must be improved across languages, cultures, modalities, and use cases.

5 stages of risk management need to be followed:

  1. Risk identification
  2. Risk assessment
  3. Risk evaluation
  4. Risk mitigation
  5. Risk Governance

The Monitoring and Intervention Methods introduced in the report are system-level safeguards that can be applied to inputs, outputs, and models to monitor AI behaviors and, when necessary, intervene:

Call to Action: Read the entire report , available ​here​. It can be your blueprint for developing strategies for your organization to adopt AI responsibly, which can play a pivotal role in business growth.

​

Weekly News & Updates…

Last week’s AI breakthroughs marked another leap forward in the tech revolution.

  1. There is a news published that OpenEuroLLM, the first family of open source Large Language Models covering all EU languages, has earned the first STEP Seal, link​
  2. OmniHuman-1 from ByteDance is released to create lifelike human videos based on a single human image. A Diffusion Transformer-based AI model generates highly realistic human videos from a single image and audio/video input. link​
  3. OpenAI launched Deep Researcher, which can do multi-step research on the internet for complex tasks. link​
  4. Tülu 3 405B, an open-source post-training model, link​

The Cloud: the backbone of the AI revolution

  • Cloud AI Resources Where You Need Them: Announcing NVIDIA L40S GPU on Oracle Compute Cloud@Customer and Oracle Private Cloud Appliance link​
  • Designing sustainable AI: A deep dive into TPU efficiency and lifecycle emissions, link​

Generative AI Use Case of the Week:

Several Generative AI use cases are documented, and you can access the library of Gen AI Use cases. Link​

Nvidia has published a survey of approximately 600 financial services professionals titled ‘State of AI in Financial Services.’​

Favorite Tip Of The Week:

Here’s my favorite resource for the week.

5-hour conversation of Lex Fridman with Dylan Patel and Nathan Lambert covering everything about AI, Generative AI, DeepSeek, Agents, etc.

​Lex Fridman Podcast​

Things to Know…

Google has published version 2.0 of the Frontier Safety Framework with the focusing of targeting:

  • To identify capability levels at which AI models without additional mitigations could pose severe risk
  • Implement methods to achieve capability levels
  • Prepare and communicate mitigation plans before such capability levels are achieved.
  • When suitable, engage outside parties to help shape and direct our approach.

The Opportunity…

Podcast:

  • This week’s Open Tech Talks episode 154 is “Generative AI Risks and Governance: What Business Leaders Need to Know with Terry Ziemniak.” he is a Fractional CISO and Partner at TechCXO

Apple | Spotify | Amazon Music ​

Courses to attend:

Events:

Tech and Tools…

  • ​Oumi is an entirely open-source platform that streamlines the entire lifecycle of foundation models, from data preparation and training to evaluation and deployment.

Data Sets…

  • ​The Tulu 3 SFT mixture​ , which contains 939,344 samples, was used to train the Tulu 3 series of models.

Other Technology News

Want to stay updated on the latest information in the field of Information Technology? Here’s what you should know:

  • ​Jump, a leading provider of artificial intelligence (AI) solutions for financial advisors and other financial services providers, today announced the successful completion of a $20 million Series A funding round. link​

And that’s a wrap!

Thank you, as always, for taking the time to read.

I would love to hear your thoughts. Please reply and share what you found most valuable this week. Your feedback means a lot to me.

Until next week,

Kashif Manzoor

The opinions expressed here are solely my conjecture based on experience, practice, and observation. They do not represent the thoughts, intentions, plans, or strategies of my current or previous employers or their clients/customers. The objective of this newsletter is to share and learn with the community.

Related posts:

  1. AI Safety and Regulation
  2. Key Risks Associated with Generative AI
  3. Exploring Global AI Trends: Insights from the AI Index 2024 Report by Stanford HAI
  4. AI Under the Guardrails of Regulations