- AI Odyssey
- Posts
- ChatGPT is lazy...π¬
ChatGPT is lazy...π¬
Plus: Why the AI Act was so hard to pass
Hello! Users are frustrated with ChatGPT's apparent reluctance to handle data-intensive tasks, even on paid accounts. OpenAI acknowledged the issue on December 7, revealing no updates for GPT-4 since November 11th.
β Lavena Xu-Johnson, Rosa Cecilia, Hubert Trinkunas
PRESENTED BY WORDTUNE
Start Your AI Journey With an All-In-One Reading and Writing Assistant
Wordtune is an AI writing assistant that helps users create compelling content. Wordtune understands the context of your writing, providing intelligent suggestions for improvement. Whether itβs enhancing sentence structure or expanding your vocabulary, Wordtune saves you time & effort.
β° Top 5 Headlines
πLink: Grimes teams up with startup on AI-equipped plush toy.
πLink: Pope Francis calls for Artificial Intelligence regulation treaty.
πLink: H&R Block launches AI tax filing assistant.
πLink: Instagram now lets you make AI-generated backdrops for your Stories.
πLink: Putin denies using body double in conversation with his own AI deepfake.
Google Cloud CEO on Gemini's value for businesses
![](https://media.beehiiv.com/cdn-cgi/image/fit=scale-down,format=auto,onerror=redirect,quality=80/uploads/asset/file/6f2f8bfc-9153-4ef8-84ab-45c4dc437e3c/1702425659342.jpg?t=1702571541)
Google Cloud CEO Thomas Kurian underscores the importance of Gemini for companies seeking to diversify beyond a single vendor, like OpenAI. Despite Google's pioneering role in generative AI research, it has been perceived as playing catch-up to rivals Microsoft and OpenAI.
Post-OpenAI drama surge: Google experienced a surge in interest and gained new customers following the recent OpenAI turmoil, where CEO Sam Altman was fired and rehired in less than a week.
Gemini models: Google announces the availability of Gemini Pro for businesses in its AI Studio and Vertex AI developer tools, with a higher-end Gemini Ultra model currently in testing with select customers.
Multimodal breakthrough: Kurian highlights Gemini's ability to handle voice, video, text, and images simultaneously, setting it apart from competitors. This includes an improvement in output for computer code, an area where Google has made strides.
Integrated advantage: Google's advantage lies not only in building models like Gemini but also in providing the cloud infrastructure for training and operating these models, resulting in cost and performance gains passed on to customers.
OpenAI comparison: While OpenAI uses Microsoft's Azure for training and hosting, Google's collaboration between model development and cloud infrastructure provides efficiencies and cost savings. The standard Gemini Pro model currently uses text for input/output, with a version accepting images as input (Gemini Pro Vision) and a fully multimodal version yet to be released.
AI is a double-edged sword for climate change
AI companies say the technology can be used to help climate change, but it is also contributing to the problem.
In the ongoing battle against climate change, AI emerges as a potential game-changer. Tech titans hype its prowess, but there's a catch β AI's carbon footprint. Google DeepMind's Sims Witherspoon admits, "AI is energy-intensive," throwing shade on the tech's environmental impact.
Energy guzzler: Witherspoon spills the beans on AI's thirst for power. Currently contributing 1-1.5% to global electricity use, AI's hunger for energy surpasses traditional computing. Brace yourself β a study warns it might gobble up as much electricity as the Netherlands by 2027.
Race for solutions: Big names like Google are on a mission to slim down AI's energy diet. Enter a carbon emissions dashboard, letting researchers tweak experiments for a greener impact. But can these efforts counterbalance AI's growing appetite?
Hope in unexpected places: Despite the eco-concerns, Witherspoon sees a silver lining. AI could be a green ally in ecology and biodiversity efforts. Think DeepMind's Serengeti project, where AI helps track animal migration β a glimpse into AI's positive potential.
Why the AI Act was so hard to pass
The EU AI Act faced a tumultuous two-year journey, culminating in a frenetic three-day debate. Policymakers grappled with chaos over the regulation of facial recognition tech and oversight of potent "foundation" models.
Foundation models and GPAI struggles:
The Act, conceived in 2021, lagged behind the evolving AI landscape, especially with models like ChatGPT. Tensions peaked over defining rules for General Purpose AI Systems (GPAI).
Countries like France, Germany, and Italy vied to exempt GPAIs, fearing innovation setbacks, injecting last-minute drama into negotiations.
Facial recognition battleground:
The European Parliament initially pushed for a full ban on facial recognition. Compromises emerged, allowing limited use under strict law enforcement conditions and delays.
Tech giants' lobbying led to a more flexible two-tier system for GPAIs, easing regulations in the final hours. The official text is pending, adding suspense to the Act's resolution.
More on AIβ¦
Deepfakes for $24 a month: how AI is disrupting Bangladesh's election.
4chan, X ease the spread of AI-generated antisemitism, racism.
News publishers see Google's AI search tool as a traffic-destroying nightmare.
![](https://www.vpdae.com/open/7133.gif?opens=1)