👐 Musk Turns Grok Loose in Open Source Saga

Hello AI Enthusiast,

Last week, we covered the news of Elon Musk suing OpenAI, saying they've lost their way from their original goal to make AI that helps everyone. He thinks they're too caught up with making money with Microsoft now.

So we wanted to know your opinion. Was Elon just trying to get attention, or was there something big going on? Well, you had a lot to say! About 40% of you think Elon's just looking for the spotlight, and another 40% are just enjoying the drama.

And now, Elon's saying he's going to share his own AI project, Grok, with everyone, making it open for anyone to see and use. He says it's to keep AI open and honest. But is it really that simple? In the news Bytes section, we give our take on that.

And now on with the news. 👇

News Bytes 🗞️

• Elon Musk's AI startup xAI is set to open source Grok, its chatbot competitor to OpenAI's ChatGPT, this week amidst a legal battle with OpenAI over alleged closed-source practices and prioritization of profits under Microsoft's ownership. The move aligns with Musk's long-standing advocacy for open-source technology and joins other firms like Meta and Mistral in releasing chatbot codes to the public.

  • 💡 Our take: Elon Musk's move follows a trend of open-sourcing AI models which usually drive profits from other sources, like user data. Meta adopted a safety-focused approach, is Elon going to do the same? AI openness isn't a one-size-fits-all solution. Ensuring ethical use should always remain a priority.

• AI company Cohere has introduced a new AI model, Command-R, designed for robust tasks like Retrieval Augmented Generation (RAG), aiming to help companies enhance customer service and automate complex workflows. It reportedly outperforms competitors like Mixtral and OpenAI's GPT-3.5 Turbo, with affordable prices and availability on major cloud platforms soon, making it a cost-effective solution for building AI applications.

  • 💡 Our take: Cohere's Command-R is a smart update that values usefulness more than being overly smart. It meets the real need for automation and making decisions. We agree that AI should solve real tasks well and real tasks, which typically don't require the assistance of a PhD-level expert.

• Cohere and Accenture collaborate to bring generative AI solutions, leveraging Cohere's models and RAG capabilities, for increased productivity and efficiency in enterprises, such as powering Accenture's knowledge agent for its Finance and Treasury teams. The partnership aims to offer customized AI solutions to businesses, addressing concerns around generative AI adoption and providing tailored models.

• The European Union's Parliament has approved the world's first major laws for controlling artificial intelligence. The EU AI Act, which sorts AI into different levels of risk, is expected to start in May after a final review and the go-ahead from the European Council, with the rules being applied starting in 2025.

• Apparently, there’s been a surge of AI-generated fake images of black people having a good time with Trump. Concerns arise over the potential for these manipulated images to influence voters, fueling worries about the spread of misinformation. Social media platforms and experts are ramping up efforts to combat the emergence of fake content, as fake images can pose a threat to the electoral process.

  • 💡 Our take: While the concern over AI-generated fake images influencing voters is valid, let's not forget the longstanding history of photo manipulation in politics. The focus on these instances distracts from the broader issue of misinformation. Instead of sensationalizing, let's address the root problem. AI is a tool that mirrors our intentions.

• Inflection AI, the company behind the personal AI chatbot Pi, recently launched a new model – Inflection 2.5. Similar to GPT-4, but with significantly lower training costs, this model powers the company's chatbot, which currently has an impressive 6 million monthly active users who engage in conversation for over 33 minutes on average and manifests a steady growth of 10% month-on-month.

  • 💡 Our take: Compared to other companies, Inflection is able to differentiate itself with its way of talking about its technology highlighting how user-friendly and personal it is, saying that people spend an average of 33 minutes interacting with it. However, this statement made us wonder about how accurate this data is. Do users really spend nearly 30 minutes actively talking to the chatbot?

• Recall a few months back when OpenAI's CEO, Sam Altman, was fired and then quickly reinstated? With a blog post, OpenAI officially puts former CEO Sam Altman back onto its board, alongside three new members. The investigation's findings clarified that the November 2023 board conflicts were not due to product or financial issues but to a difference in perspective, marking an end to recent turbulence.

• At the same time, a New York Times investigation disclosed that this unexpected turn of events was due to allegations of manipulative conduct made by CTO, Mira Murati, and co-founder, Ilya Sutskever.

  • 💡 Our take: OpenAI's recent leadership issues have shown the personal conflicts and drama behind the technology, making it clear there's more going on than just tech work. This situation helps the New York Times' lawsuit against OpenAI by revealing that OpenAI is dealing with a lot of chaos. It points out how important it is for tech companies to be honest and trusted.

• Two former DeepMind employees have launched Haiper, an AI-powered video-generation tool allowing users to create short videos for free. The company aims to solve the uncanny valley problem and plans to offer its core video-generation model to others. After OpenAI’s Sora release the battle is heating up.

• China is testing an AI assistant for neurosurgeons called CARES Copilot 1.0, developed by a Hong Kong-based agency under the Chinese Academy of Sciences. The AI model is trained to offer surgical consultations. It’s always amazing to see AI put to good use.

Educational Pill 💊

Choosing the right AI model to power your product

We can’t stress this concept enough. Choosing the right AI model for your product hinges more on speed and practical utility than on high-level intelligence metrics. High-profile models like GPT-4, Gemini Ultra, or Claude 3 Opus are celebrated for their prowess in complex domains such as math or science.

However, for everyday applications—like assisting with blog writing, text summarization, or data analysis—such sophistication isn’t just unnecessary; it’s inefficient. The real value lies in an AI's ability to deliver quick, accurate, and cost-effective solutions. Most tasks don’t demand Einstein-level intellect but rather a reliable, fast-responding AI that enhances user experience without inflating costs. As businesses look to scale, prioritizing efficiency ensures that AI products can grow without unsustainable expenses.

That's a wrap on our newsletter! Here’s a quick recap before you go:

• Free Webinar on March 15th at 6 pm CET: Learn about Generative AI Implementation Strategy.

• Free Webinar on March 22nd at 6 pm CET: Discover how we at AI Academy use AI.

• Empower Your Team: If you're a startup or SME, reach out for a personalized AI course offer.

• Customized Corporate Training: We offer tailored sessions for companies diving into AI.

Catch you next week! 👋