OpenAI Drama 🤯 Google's AI Fails 🙊 and the Rise of Elon Musk's xAI
Your weekly dose of AI insights: stay ahead of the curve and understand the future of technology.
This Week in AI #12
Every week, we try to demystify this world-changing technology.
Check out the most interesting events below and hopefully, you’ll learn something new.
At a glance:
Details about Sam Altman’s (OpenAI’s CEO) firing last year have come to light.
Google’s search AI features display wildly inappropriate content.
Musk’s xAI startup just became the second most valuable AI company ($24 billion valuation) in a year.
Current LLM architecture won't reach human-level intelligence
Canva Create has seriously good AI features
GPT-4 is better at picking stocks than humans
👋 If you’re new here, welcome!
Subscribe to get your AI insights every Thursday.
Out in the Open: Altman’s Ousting Last Year
In November 2023, OpenAI's board unexpectedly fired CEO Sam Altman despite his success with ChatGPT.
Former board member Helen Toner reveals that the decision stemmed from a loss of trust in Altman due to his lack of transparency, alleged misconduct, and a toxic work environment.
Altman was quickly reinstated due to pressure from employees, Microsoft, and a fear of the company's collapse without him.
Why It Matters:
This episode is just the latest in a string of controversies surrounding OpenAI.
The company has recently faced public scrutiny over issues such as the departure of its safety team, concerns about the source of its training data, and a dispute with Scarlett Johansson over the use of her voice in an AI model.
Ironically, Johansson played the voice of an AI assistant in the 2013 romance sci-fi movie “Her.” Beginning in September, Altman reportedly attempted to court the actress but was turned down on multiple occasions.
Despite these challenges, OpenAI continues to push the boundaries of AI development, with their new model GPT-4o driving record revenue and the company already training its next flagship model.
Google AI Features Fall Flat, Again…
Google's new AI search feature, AI Overviews, is facing backlash due to its tendency to generate inaccurate, and sometimes dangerous responses.
The technology's reliance on popularity rather than truth and its susceptibility to biases and misinformation present significant risks for Google's reputation and its most lucrative service, search.
Why It Matters:
The issues with AI Overviews highlight the challenges and potential dangers of deploying AI tools that generate content at scale.
The erosion of trust in Google Search and the potential for an internet flooded with low-quality, AI-generated content raises concerns about a negative feedback loop, where results degrade over time.
xAI Rockets to a US$24Bn Valuation
Elon Musk's AI startup, xAI, has raised $6 billion in funding, bringing its valuation to US$24 billion ($36.3 billion). This makes xAI the second most valuable AI startup globally, behind OpenAI’s US$80 billion ($121 billion) valuation.
Even though it’s only a year old, xAI has garnered significant financial backing from prominent investors. Funding will be used to develop xAI's products and accelerate its research agenda, according to the company.
Why It Matters:
Substantial funding and rapid growth of xAI signals a possible shift in the AI industry.
With Elon Musk's ambitious goals for xAI and his ongoing rivalry with Altman’s project, the development could intensify competition and further drive innovation in the technology.
Musk's recent commitment to releasing AI models that surpass open-source alternatives is a huge win for the open-source community.
Current AI Won’t Reach Human-level Intelligence
Meta's AI chief Yann LeCun believes that Large Language Models like ChatGPT will never achieve human-level intelligence.
LeCun believes limitations for such models stem from the inability to understand the physical world — impacting the tech’s reasoning and planning.
He proposes an alternative approach called "world modelling," where AI systems learn about the world like humans do.
Why It Matters:
LeCun's perspective challenges the prevailing narrative around LLMs and offers a different vision for the future of AI.
His emphasis on world modelling could lead to the development of more sophisticated and safer AI systems.
The approach faces significant challenges and scepticism from some experts, adding to existing fears about the technology’s development in the wild.
In any case, the outcome of Meta's investment in both LLMs and LeCun's vision could have significant implications for the company's future and the broader AI landscape.
Canva Becomes a Serious Contender With Latest AI Features, Released At “Cringe” Event
Canva Create 2024 introduced significant updates to the design platform, including new AI-powered features and Canva Enterprise for businesses.
The event's promotional rap video went viral, drawing attention to the new features while sparking debate about its unconventional style.
Canva's co-founder embraced the "cringe," highlighting the company's unique culture and ability to generate buzz. Watch the rap.
Why It Matters:
Canva's continued growth, with over 170 million monthly users and a valuation of $58.9 billion, is yet another example of a major company exploring the tech.
Founded in Perth, Australia, Canva has become one of the country’s most successful software companies since launching over a decade ago.
Attempting to maintain that momentum, Canva’s AI features are going after established creative production tools, including Adobe Photoshop and After Effects.
Study: AI Model GPT-4 Beats Human Investors in Picking Stocks
This study examines the ability of a large language model (GPT-4) to analyse financial statements and predict future earnings changes. The model outperforms human analysts, particularly in situations where analysts typically struggle.
Its predictive accuracy is on par with a state-of-the-art machine learning model specifically trained for this task. The LLM's success stems not from memory but from its ability to generate valuable narrative insights from financial data.
Why It Matters:
The findings suggest that LLMs like GPT-4 can revolutionise financial analysis by offering a powerful tool for interpreting financial statements and predicting earnings changes.
This could democratise access to sophisticated financial analysis, possibly benefiting both individual investors and institutions.
The study also highlights the importance of narrative insights in financial analysis, suggesting that LLMs can complement and enhance human decision-making.
That’s a Wrap!
If you want to chat about the things I wrote, you can reach me through LinkedIn.
Or give my editor a bell through his LinkedIn here.
If you liked it, give it a share!