Google’s DeepMind’s Gemini 2.5 Raises the Bar
Google’s Gemini 2.5, featuring its new Deep Think, puts the company further ahead and in a stronger position against OpenAI and Anthropic for large-scale, enterprise-grade LLM Deployments. More information can be found here.
Perspective: I am using both Gemini and OpenAI and find they both often give different answers, but with a critical eye, you have more insights to help advance knowledge worker productivity. Fact checking is critical and high-risk stakes as well, as recently found the hard way in legal court, where the judge found that the case arguments were not factual and sourced from LLMs, making up false precedents. Immediate credibility impact. Deep Think must be all about trusted Data sources.
USA Senators Reject A 10-Year Ban On State-Level AI Regulation Is A Big Blow To The Tech Industry
US Senators voted on July 1 to remove a controversial 10-year moratorium on state regulation of artificial intelligence from President Trump’s “One Big Beautiful Bill,” marking a significant defeat for a tech industry that had lobbied hard to keep the provision in the sweeping tax and spending package. Lawmakers voted 99-1 in an overnight session to remove the provision by adopting an amendment tabled by Marsha Blackburn, Republican of Tennessee, who had earlier broken with her party over the issue. Companies such as OpenAI and Google had previously argued in support of blocking states from regulating AI, to avoid what they said would be a patchwork of rules that could hamper innovation.
Perspective. This was the right decision, as we have too many gaps in corporate purpose in aligning AI to ensure human safety and security in high-risk AI applications. Did we allow the airline industry to evolve without checks and balances to ensure human safety? No, we worked on international laws that protected travel in the skies. As we advance Agentic AI into more data streams, we need to ensure we have accountabilities and humans in the loops on high-risk AI applications.
Goldman Sachs Advances AI Intelligent Agents
Goldman Sachs has just rolled out AI tools that could replace a legion of spreadsheet jockeys (a.k.a. junior bankers). Goldman Sachs has 10,000 people currently using the new GS AI Assistant, ‘The first generative AI-powered tool to reach this scale,” says Goldman Sachs CIO Marco Argenti. If this move cuts the ridiculously usurious banker fees for entrepreneurial companies (7 percent off the top for an IPO—yikes!), we are all for it.
The GS AI Assistant enhances productivity by assisting with tasks such as summarizing complex documents, drafting communications, analyzing data, coding, and translating research into multiple languages. It integrates with various large language models (LLMs), including OpenAI’s GPT-4o, Google’s Gemini, and Claude 3.7 Sonnet, operating within Goldman’s secure compliance framework to ensure data privacy. This is a perfect case study of how AI will reshape the white-collar world.
More information can be found here.
Perspective: This is the beginning of more worker reductions unless we think hard about how jobs are going to shift into new roles and capabilities. The biggest risk we have is not thinking of the consequences and preparing for our workforce evolution. We need deeper change management planning at the front of the AI agent assistants, vs saying oops and facing backlash due to a lack of corporate purpose and thoughtful planning. #PeopleReallyMatterinAI
EU AI ACT – In Final Release Stages
The European Union’s AI Act officially took effect on 1 August 2024. Key implementation dates are:
- Feb 2025: AI bans and transparency requirements begin
- Aug 2025: Rules for governance and obligations kick in
- Aug 2026: High-risk AI system requirements apply
- Aug 2027: Full compliance required for general-purpose models. Boards with operations or data touchpoints in Europe should be planning now.
- See more information here.
Perspective: Having safety guard rails defining high-risk applications that could cause humans harm is very wise. I wish Canada and the USA could align faster on classifying AI high-risk risks and look at specific AI use cases, and avoid generalities. High-risk applications are simply thinking deeply about the AI use case and all the surrounding context risks.
- For example, would we want a world where we could social score the value of humans on every transaction they make – Surveillance Capitalism at its finest?
- Would we want a world where an AI virus could take over all the utility infrastructure and no longer provide power and not be able to have a kill switch and an alternative control path?
- Do we want to have a robot operate on us and have no means to have humans in the loop to ensure our safety?
These are just some examples of high-risk applications. As we mature in AI understanding, we will have AI assess the risks of the AI use cases. One crucial fact remains: we have to make choices – AI Freedom or AI balanced controls that value both innovation and human safety.
Europe is demonstrating courage to legislate some operating guardrails, and by starting to test adherence within an operating construct with consequences, it creates a learning context to adjust as knowledge is gained.
I hope we can all lead with facts vs rhetoric or naivety – we have a lot at stake if we don’t get this right. So many countries are not tackling the hard questions with a tenacity to get the job done and iterate. Learn and Fail Fast – staying in limbo land just exacerbates the loss of wisdom.
To stay updated about the latest in AI and its impact, subscribe to our CEO’s weekly AI Insights newsletter
