‘Where we are today in biology AI is similar to GPT in 2020’

InstaDeep, an African AI startup acquired by BioNTech for over $550 million in 2023, has made substantial progress in integrating AI into the biotech company's pipeline. Operating under the German biotech's umbrella while maintaining operational independence, InstaDeep aims to enhance drug discovery and therapeutic development with advanced AI capabilities. Over the past year, InstaDeep has worked on several impactful projects, including accelerating histology analysis through visual AI tools and contributing to BioNTech's mRNA-encoded antibody technology, which is now being used in cancer immunotherapy.

InstaDeep recently unveiled the Bayesian Flow Networks (BFN), a generative AI model for proteins that outperforms existing models in producing natural and functional proteins. This development is part of a broader effort to deploy AI to boost BioNTech's research capabilities, particularly in the fields of genomics and protein engineering. InstaDeep also introduced a new near-exascale supercomputer, positioning its partnership with BioNTech as one of the leaders in computing power globally.

Despite the focus on biotech, InstaDeep continues to work on industrial optimization projects independently. Notable initiatives include optimizing railway planning for Deutsche Bahn and improving airport operations for Fraport. Additionally, the company recently launched DeepPCB Pro, an AI-assisted tool for printed circuit board design. InstaDeep's continued efforts in both biotech and industrial sectors have allowed it to grow significantly, with new offices and expanded geospatial intelligence capabilities in Africa, such as using satellite imagery to predict locust breeding grounds, supporting food security efforts across the continent.

Free Notion and Unlimited AI

Thousands of startups use Notion as a connected workspace to create and share docs, take notes, manage projects, and organize knowledge—all in one place. We’re partnering with Notion to offer 6 months of new Plus plans, including unlimited Notion AI so you can try it all for free!

To redeem the Notion for Startups offer:

1. Submit an application using our custom link: https://ntn.so/superpowerdaily and select Superpower Daily on the partner list.

2. Include our partner key, STARTUP4110P37060.

Billions of Gmail users at risk from sophisticated new AI hack — how to stay safe

Hackers have launched a new sophisticated phishing attack targeting Gmail users, with tactics that combine both spoofed communication and AI-driven calls. Sam Mitrovic, a Microsoft solutions consultant, recently reported his experience of being targeted by such an attack. It started when he received a Gmail account recovery notification, which he denied, followed by a missed call from a Google Sydney number. A week later, he received another account recovery request, and this time, a supposed Google Support call followed. Though the call appeared convincing, Mitrovic noticed suspicious details, such as an unusual email address and a seemingly perfect AI voice on the call, which ultimately tipped him off to the scam.

The hackers used AI-generated voices and fake Google phone numbers to trick users into giving away their Gmail credentials. Mitrovic's cautious approach and vigilance helped him identify the scam, but he warned that the phishing attempts are getting increasingly sophisticated, designed to deceive even experienced users. He urged individuals to remain vigilant, scrutinize communications for discrepancies, and seek help from trustworthy sources if uncertain about a situation.

Meanwhile, Google has joined forces with the Global Anti-Scam Alliance (GASA) and DNS Research Federation to launch the Global Signal Exchange, an intelligence-sharing initiative designed to combat online scams. This new platform, powered by Google Cloud, allows participants to share real-time insights into cyber threats and leverage AI to detect patterns and disrupt scams more effectively. Users are reminded to practice caution by checking sender details, avoiding clicking on suspicious links, and installing reliable antivirus software to protect against cyber threats.

This AI Pioneer Thinks AI Is Dumber Than a Cat

Yann LeCun, a prominent AI researcher and NYU professor, is challenging the growing narrative that artificial intelligence is approaching human-level intelligence and poses existential threats. LeCun, who helped pioneer the AI boom and is known as one of the "godfathers" of the field, believes that much of the hype surrounding AI's power and danger is overstated. He argues that today's AI systems are far from surpassing human intelligence—in fact, they are not even as intelligent as a cat. LeCun's skepticism stands in contrast to other AI experts like Elon Musk and Geoffrey Hinton, who have warned about AI's potential dangers.

LeCun's criticism focuses on the limitations of current AI models, such as large language models (LLMs) that drive technologies like ChatGPT. He asserts that these systems, while powerful, are merely predicting the next word in a sequence and lack genuine reasoning abilities or common sense. According to LeCun, the impressive language capabilities of AI models often fool people into thinking they are more intelligent than they actually are, but these models are simply sophisticated parrots, not thinkers.

At Meta, where he serves as the chief AI scientist, LeCun is advocating for a different approach to developing AI. He believes that future advances in AI will require systems that learn in ways similar to how animals or humans do—for example, by building models of the world from visual data. This approach, he says, could pave the way to true artificial general intelligence (AGI), but it is still decades away. LeCun remains optimistic about AI's potential but insists that today’s methods are not enough to reach human-level intelligence, calling for innovative approaches beyond simply scaling up current technologies.

Other stuff

• Terence Tao, mathematician: ‘It’s not good for something as important as AI to be a monopoly held by one or two companies’

• Integer addition algorithm could reduce the energy needs of AI by 95%

• Artificial intelligence helps Aussie farmers target weeds, livestock illnesses and pests

• Tesla unveils Robotaxi

• Anthropic CEO goes full techno-optimist in 15,000-word paean to AI

• A profile of Ben Nimmo, OpenAI's principal threat investigator who found evidence that Russia and China were using ChatGPT to sway political discourse online news

• Some of the coolest examples of OpenAI Real-time API

Hope you enjoyed today's newsletter

⚡️ Join over 200,000 people using the Superpower ChatGPT extension on Chrome and Firefox.