- Supercharged AI
- Posts
- ⚡️ ChatGPT Who?
⚡️ ChatGPT Who?
PLUS: Google's AI model reveals disease-causing genes
Good morning. Google's conversational AI, Bard, just got a massive upgrade. But will it be a helpful assistant or just another way for Big Tech to peek into your life? Let's find out.
Today’s Highlights:
OpenAI formalizes Red Teaming Network for better AI models
Chan Zuckerberg Initiative funds high-end GPU cluster
Transparency in AI content with TikTok's new tools
DEEP DIVE
Google Bard’s Latest Upgrade

Google announced a significant upgrade that extends Bard's capabilities to some of the most used productivity apps—enhancing day-to-day tasks with an even stronger emphasis on accuracy.
Bard Extensions and App Integration
“We wanted to make sure that the way that we bring this to users is extremely rooted in the three principles that we have...the trust that we build with our with people that use Bard, which is around transparency, choice, and control.”
Following these principles, Bard will only interact with apps such as Gmail, Drive, and Docs if the user opts in for it, and they can revoke this permission at any time. Further, the data accessed during such interactions will not be used for reinforcement learning or seen by human reviewers.
Imagine this: Bard could be asked to summarize your missed emails or retrieve dates from an email thread and use them to pull up real-time flight and hotel information for a planned trip. It would provide map directions and even pull up YouTube videos of recommended activities at your destination!
However, Bard is still a learner, so don't expect perfect accuracy. For example, it won't be able to find your wine club memberships unless the term "wine club" is explicitly mentioned in an email.
Double-Checking Answers and Collaborations

The "Google it" button feature of Bard has been improved, allowing the chatbot to double-check answers using Google search validation.
Additionally, Bard users can now collaborate easily. If a Bard chat is shared through a public link, anyone can continue the conversation and ask Bard further questions.
Not only that, Google is also expanding the availability of Bard’s existing features in over 40 new languages.
PUNCHLINES
Nvidia’s Got Competition: SambaNova reveals its latest chip, SN40L, designed to handle a 5 trillion parameter model.
Meteor Shower: Intel aims to democratize AI with its new processor, promising to shake up the "Siliconomy".
Sanity Check: Galileo's new toolset will help you understand why your AI model is hallucinating.
New Heights: Pyron raises $100M to accurately index and analyze enterprise data.
In Check: TikTok debuts new tools to label AI-generated content for transparency.
TLDR
OpenAI establishes Red Teaming Network for more robust AI models: OpenAI introduces the Red Teaming Network to enhance the evaluation and risk mitigation of its AI models. The network solidifies current collaborations with external experts, allowing deeper exchanges with various scientists, research organizations, and civil society groups. OpenAI invites worldwide domain experts to apply, prioritizing geographic and domain diversity.
Chan Zuckerberg Initiative builds GPU cluster for medical breakthroughs: The Chan Zuckerberg Initiative is funding a high-end GPU cluster to generate predictive models of healthy and diseased cells using AI, aiming to assist in curing, preventing, or managing all diseases by the century's end. The "openly available" models will be accessible to researchers and will be trained using multiple datasets, with the goal of creating a "virtual biology simulator."
Digimarc offers copyright protection for AI training data: Digimarc's new service, Digimarc Validate, allows copyright owners to insert ownership identification in metadata enabling AI models to recognize copyrighted material. The software company suggests this will improve the treatment of copyrights in AI training data. The service is now available starting at $399/mo.
Google's AlphaMissense reveals disease-causing genes: Google's AI model, AlphaMissense can classify 89% of all possible “missense” variants in genes compared to the 0.1% classified by human experts. AlphaMissense will help researchers preview thousands of proteins at a time and focus on deciding where to prioritize. Google will make the code freely available to the research community.
TRENDING TOOLS
📚 Einblick: Manage your data, code-fixing, and chart plotting effortlessly
🤖 Astrocyte: Interact and chat with 3D AI characters
💼 Kick: Streamline your bookkeeping with AI-powered tools
💻 CodeWiz: Handle coding challenges effortlessly with AI assistance
🔍 Simon: LLM chat and semantic search with Postgres and 10 lines of code
That’s all for today—if you have any questions or something interesting to share, please reply to this email. We’d love to hear from you!
P.S. If you want to sign up for the Supercharged newsletter or share it with a friend, you can find us here.
Reply