Daily AI Newsletter: OpenAI’s Sora will be publicly available later this year

In today’s email:

☠️ Hackers can read private AI-assistant chats even though they’re encrypted
👀 Nvidia offers developers a peek at a new AI chip next week
📚 What I learned from looking at the 900 most popular open-source AI tools
🧰 11 new AI-powered tools and resources. Make sure to check the online version for the full list of tools.

OpenAI’s Sora text-to-video generator will be publicly available later this year

OpenAI is set to make its innovative text-to-video generator, Sora, available to the public later this year, as revealed by the company's CTO, Mira Murati, in an interview with The Wall Street Journal. Initially offered to a select group of visual artists, designers, and filmmakers, Sora has the capability to create hyperrealistic videos from text prompts. Future updates will include the integration of audio to enhance realism and tools allowing users to edit the AI-generated content.

Murati remained vague about the specific data used to train Sora, mentioning only that it included publicly available or licensed data, with a confirmed partnership with Shutterstock. The technology, while more costly to operate than OpenAI's text-to-image model DALL-E, aims to be released at a comparable price point. Examples of Sora's potential include generating animated videos like a bull in a China shop and a mermaid reviewing smartphones.

As the 2024 presidential election nears, there are rising concerns about the potential misuse of generative AI technologies like Sora in spreading misinformation. To address this, OpenAI plans to implement restrictions, such as prohibiting the creation of videos with public figures and adding watermarks to distinguish AI-generated content from real footage, though the effectiveness of watermarks as a deterrent remains under scrutiny.

AI brews beer and your big ideas

What’s your biggest business challenge? Don’t worry about wording it perfectly or describing it just right. Brain dump your description into AE Studio’s new tool and AI will help you solve that work puzzle.

Describe your challenge in three quick questions. Then AI churns out solutions customized to you.

AE Studio exists to solve business problems. They build great products and create custom software, AI and BCI solutions. And they once brewed beer by training AI to instruct a brewmeister and then to market the result. The beer sold out – true story.

Beyond beer, AE Studio’s data scientists, designers and developers have done even more impressive things working 1:1 with founders and executives. They’re a great match for leaders wanting to incorporate AI and just generally deliver outstanding products built with the latest tools and tech.

If you’re done guessing how to solve work problems or have a crazy idea in your back pocket to test out, ask AI Ideas by AE Studio for free solutions, right now.

Hackers can read private AI-assistant chats even though they’re encrypted

Researchers have discovered a new side-channel attack that can decrypt the responses of AI assistants, such as ChatGPT, despite their encrypted communications. This vulnerability, which doesn't affect Google's Gemini, allows attackers to deduce the content of AI assistant interactions with a surprising degree of accuracy. By monitoring the size and sequence of encrypted data packets, the attackers can infer the topics and details of the conversations, posing significant privacy risks.

The attack exploits the way AI assistants transmit tokens, which are units of text, in real time, revealing a pattern that can be analyzed to reconstruct the original message. Even though the responses are encrypted, the sequential transmission of tokens gives away information about their length, enabling attackers to piece together the messages with the help of specially trained large language models (LLMs). This vulnerability has been demonstrated to accurately predict the content of conversations in various scenarios, raising concerns about the confidentiality of AI-assisted communications.

In response to this discovery, some AI service providers, including OpenAI and Cloudflare, have started implementing countermeasures such as packet padding to obscure the lengths of the tokens. These steps aim to mitigate the risks posed by this side-channel attack, but they may also affect the user experience by introducing delays or increasing data traffic. The research underscores the ongoing challenge of securing AI-powered communication against sophisticated eavesdropping techniques and highlights the need for continuous vigilance and innovation in cybersecurity practices.

Nvidia offers developers a peek at a new AI chip next week

Nvidia is set to unveil its latest AI processor, expected to be named the B100, at its annual developer conference, the first in-person event since the pandemic. With a market capitalization nearing $2 trillion, Nvidia's revenue is anticipated to jump 81% to $110 billion this year, driven by high demand for its AI chips used in various technologies. Despite a significant lead in AI computing, the challenge for Nvidia is maintaining its dominance in the evolving tech landscape.

The B100 chip is at the heart of Nvidia's strategy to dominate AI computing, expected to surpass the performance of its predecessors and support a range of AI applications. Nvidia's stock has seen an impressive 83% increase in 2024, reflecting the high expectations from its financial performance and upcoming products. However, analysts caution that the company's valuation could be at risk if these high expectations are not met.

Nvidia's global position is also influenced by geopolitical dynamics, particularly its competition with Chinese firms in the AI chip market. Despite restrictions on China's access to Nvidia's top-tier chips, Chinese companies have not yet matched Nvidia's latest offerings. The B100 is anticipated to further Nvidia's lead, underscoring the rapid pace of innovation and the competitive edge it aims to maintain in the semiconductor industry.