🤓Israel's AI Strategy

Israel steps up AI game with new supercomputer initiative

Key Points 

• Israel is launching a tender for its first supercomputer to strengthen its global leadership in AI.

• The government also plans to invest $250 million in a national AI program.

👨🏻‍💻News - Israel is set to launch a tender next month to establish its first supercomputer, aimed at maintaining the country's position as a global leader in AI. 

Dror Bin, CEO of the Israel Innovation Authority, announced this initiative at an AI conference, emphasizing the importance of staying ahead in AI technology. He highlighted the dual nature of AI as both an asset and a potential challenge, underscoring the need for proactive measures. 

Bin also revealed that the government plans to invest $250 million in a national AI program, with significant contributions from government, industry, and academia. Sixty percent of this budget is earmarked for implementation by 2024, with the program expected to conclude in 2027, potentially with additional funding.

🤓Why this matters - Tech makes up 20% of Israel's economic output, solidifying its reputation as one of the world's premier tech hubs. Bin noted that out of Israel's 9,000 startups, over 2,200 use AI, positioning Israel third globally with 73 leading AI firms.

Bin stressed the importance of ensuring that Israel maintains its leadership and position in the global AI race. He underscored the significance of a supercomputer in training advanced artificial intelligence models to achieve this goal.

Bin explained that when a high-tech company or researcher needs to train a large model, they have to purchase cloud computing time because there isn't a local data center equipped with enough GPUs to handle these tasks. He added that the upcoming supercomputer will help address this issue, as it will be accessible to researchers and companies at a cost below market rates.

UK think tank calls for AI incident reporting system

Key Points 

• The CLTR recommended that the next government establish a logging system for AI incidents across the UK.

• Since 2014, news outlets have recorded 10,000 AI-related safety incidents.

• Implementing an incident reporting system would thus provide rapid insights into AI malfunctions and help the government foresee similar future incidents.

☕News - A UK think tank has urged the creation of a system to record misuse and malfunctions involving AI technology. 

The Centre for Long-Term Resilience (CLTR) recommended that the next government establish a logging system for AI incidents in public services and consider a central hub to collate AI-related episodes across the UK. They emphasized that a reporting regime similar to the Air Accidents Investigation Branch (AAIB) was essential for the effective and safe use of AI technology. 

📒A case in point - The report highlighted that since 2014, news outlets have recorded 10,000 AI "safety incidents," compiled in a database by the Organisation for Economic Co-operation and Development (OECD). 

The OECD defines harmful AI incidents as those causing physical, economic, reputational, or psychological harm. Examples include a deepfake of Labour leader Keir Starmer being abusive, Google's Gemini model depicting WWII German soldiers as people of color, self-driving car incidents, and a man who planned to assassinate the late queen being encouraged by a chatbot.

🤠Potential benefits -

• The CLTR noted that many AI incidents may not be covered by existing UK watchdogs, as there is no regulator for advanced AI systems like chatbots and image generators. Implementing an incident reporting system would thus provide rapid insights into AI malfunctions and help the government foresee similar incidents in the future.

• It would also enable coordinated responses to serious incidents and identify early signs of potential large-scale harms. 

• While some AI models might only reveal issues after full release, incident reporting would allow the government to evaluate how well the country's regulatory framework is addressing these risks.

AI outperforms real students in university exams, study finds

Key Points 

• ChatGPT was used to generate exam answers for 33 fictitious students.

• The AI-generated answers averaged half a grade higher than those of real students and were often unnoticed by markers.

• AI excelled in first and second-year modules but struggled in the third year, where abstract reasoning is more crucial. 

👨🏻‍🏫News - A study by the University of Reading has found that AI can outperform real students in university exams.

Researchers used ChatGPT to generate answers for 33 fictitious students in an undergraduate psychology program. The AI-generated answers were, on average, half a grade better than those of real students and went mostly unnoticed by markers. 

🕵🏻‍♂️What's more?

• Surprisingly, 94% of AI-generated essays did not raise any concerns with markers, making them nearly undetectable. 

• AI-generated answers were submitted for first, second, and third-year modules. The AI outperformed real students in the first two years but not the third, aligning with researchers' observations that current AI struggles with the abstract reasoning required for third-year exams.

🥸In conclusion - These findings prompt educators to rethink assessment design and methods in the AI era. As AI progresses, the education sector must adapt to maintain the integrity and fairness of academic evaluations.