Empowering Europe: Google's Initiative to Skill Up Workers in AI

Google’s initiative will aim to boost the skill training of Europeans in the realm of Artificial Intelligence, and equip people with the needed skills to fully embrace AI opportunities.

Google has launched the AI Opportunity Initiative for Europe with a 25 million euros fund to enhance AI skills among Europeans, focusing on collaborations with EU governments, academics, and businesses to provide advanced AI training. Meanwhile, advancements in AI application within operating systems have been made, with Microsoft Research and Peking University overcoming challenges through novel training environments, improving AI model accuracy in OS navigation. Additionally, a JPMorgan survey reveals a major shift in the financial sector's perception, with a majority of institutional traders viewing AI and ML as pivotal technologies for the future of trading.

Google Launches 25 Million Euro AI Training Fund in Europe

Google has unveiled a new initiative, the AI Opportunity Initiative for Europe, with a commitment of 25 million euros ($26.9 million) aimed at boosting skill training for Europeans in the realm of artificial intelligence (AI). The initiative plans to equip people with the necessary skills to leverage AI opportunities, especially considering Europe is seen as a potential leader in utilizing AI for economic advancement.

The initiative is set to collaborate with European Union governments, civil society, academics, and businesses to offer advanced AI training, particularly focusing on local startups and vulnerable communities. A large portion of the fund, approximately 10 million euros, is dedicated to helping workers acquire skills to prevent them from being marginalized by the rapid advancements in technology. This is very similar to an effort by the Italian government in mid-2023, which allocated millions towards improving digital skills among workers facing job risks because of automation and AI.

Reflecting on its previous success with the "Grow with Google" program that was launched in 2015, which provided free training to bridge the digital skills gap in the EU and trained over 12 million people, Google now aims to replicate this impact in the AI domain. The initiative also includes a partnership with the Centre for Public Impact (CPI), inviting applications from EU-based social enterprises and nonprofits to maximize the outreach of AI training. The CPI's Executive Director, Adrian Brown, acknowledges the transformative potential of AI, while also acknowledging the risk it poses in widening inequality gaps.

Google is also expanding the languages available for its AI foundational course to 18 and enhancing its Google Career Certificates program with more resources for hands-on AI application in professional settings.

This initiative arrives at a critical time as EU regulators are finalizing the EU AI Act, which will set the framework for the use and development of AI technologies in the EU.

Conquering the Challenges of Operating Systems

In other AI news, scientists from Microsoft Research and Peking University have made great strides in enabling AI large language models (LLMs) like GPT-4 to function in operating systems (OS). Despite the ease of using LLMs in generating text for things like email drafting or poetry writing, their application as autonomous agents within OS environments has been quite a challenge.

Traditionally, AI models have been trained using reinforcement learning in virtual environments, mostly through modified video games like Super Mario Bros. and Minecraft, to teach concepts like self-guided exploration and goal seeking. However, the complexity of operating systems presents a very unique set of challenges. These systems require multimodal interactions involving various components, programs, and applications, making them a way different playground for AI agents.

The inherent risks of trial and error in an OS, like data loss from incorrect password attempts or forgotten shortcuts, increases the difficulty. To investigate these challenges, the team tested various LLMs, including Meta’s Llama2 70B and OpenAI’s GPT-3.5 and GPT-4, finding that their performance fell short in the OS environment. The research highlighted three primary obstacles: the vast and dynamic action space, the need for inter-application cooperation, and the necessity for solutions that align with user constraints like security and preferences.

To tackle these issues, the researchers developed AndroidArena, a novel training environment mimicking the Android OS, to better understand and enhance LLM capabilities. Their analysis pinpointed a lack of understanding, reasoning, exploration, and reflection as key shortcomings.

Remarkably, during their study, they developed a simple yet effective way to improve model accuracy by 27% by using prompts. By automatically including information about previous attempts and actions within these prompts, they essentially embedded a form of memory, addressing the critical need for reflection in AI agents.

This breakthrough not only sheds light on why LLMs struggle with operating systems but also opens up completely new pathways for integrating AI more seamlessly into complex, real-world environments. Now, the implications of this research could revolutionize how AI models interact with and navigate through operating systems.

The Future of Trading According to a JPMorgan Survey

AI is also making its mark in the financial sector. A recent survey conducted by JPMorgan observed a major shift in the attitudes of institutional investors towards the technologies shaping the future of trading. The survey, which included responses from 4,010 institutional traders across 65 countries, revealed a growing consensus on the pivotal role of AI and machine learning (ML) in trading.

61% of respondents anticipate AI and ML to be the most impactful technologies in the trading realm over the next three years. This is a pretty big increase from only 25% two years ago.

After AI and ML in the survey's technology impact rankings were application programming interface (API) integration, chosen by 13% of respondents, and blockchain or distributed ledger technology, along with quantum computing, each amounting 7%.

Meanwhile, mobile trading applications and natural language processing were selected by 6% of the participants, suggesting a degree of skepticism and caution towards these technologies. Notably, blockchain and mobile trading applications have seen a decline in their perceived promise for the future of trading, losing 18% and 23% of investor interest since 2022, respectively.

This skepticism seems to extend to cryptocurrency trading as well. The survey results showed a clear reluctance among institutional traders to engage with cryptocurrencies like Bitcoin (BTC), with 78% indicating no plans to trade digital coins within the next five years—an increase from 72% last year. However, there's a slight uptick in the percentage of respondents actively trading or starting to trade crypto, moving from 8% in 2023 to 9% in 2024.

The increasing reliance on AI and ML reflects a broader trend in the finance sector, where these technologies are credited with enhancing trade predictions and identifying real-time market sentiment threats. A report by Nvidia in 2022 highlighted the integration of AI and ML by investors, with 30% reporting a reduction in their annual revenue by more than 10% through the use of these technologies.

JPMorgan's approach to cryptocurrencies has been mixed, with CEO Jamie Dimon very publicly criticizing Bitcoin even as the company engages with the crypto market, including being named an authorized participant in a Bitcoin exchange-traded fund (ETF) by BlackRock.