ChatGPT Now Runs In Cars

PLUS: MIT's HiP uses multiple AI models for transparent, adaptable robot planning.

Good morning! Today is Tuesday, January 9th and Volkswagen adds ChatGPT to cars, generative AI threatens KYC protocols, and alternatives like cryptography are proposed. New to The Intelligence Age? Sign up here.
Drop us a line anytime at [email protected] with feedback.

News & Insights

Volkswagen Is Bringing ChatGPT Into Its Cars And SUVs

Volkswagen is steering into the future, integrating ChatGPT with its IDA system across a range of models, including their electric and conventional line-up. By leveraging Cerence's Chat Pro and OpenAI's LLM technology, the German automaker aims to enhance in-car assistance, offering drivers a more enriched and intuitive conversational experience. While initially rolling out in Europe, plans for U.S. models are pending internal approvals. This move signals Volkswagen's commitment to pushing the boundaries of vehicular AI interaction, setting a new benchmark for the industry.

Cerence's tailored LLM, CaLLM, is engineered to handle vehicle-specific queries, ensuring responses stay relevant and appropriate. Despite limitations to prevent responses on sensitive topics, clever phrasing can still elicit generic answers, as evidenced at CES 2024. This integration propels Volkswagen's IDA from a simple command-based tool to a sophisticated, conversational partner, capable of fielding general knowledge questions and delivering vehicle-specific information—all hands-free. techcrunch

GenAI Could Make KYC Effectively Useless

KYC protocols are under threat from generative AI advancements that simplify the creation of deepfake identities. Recently highlighted vulnerabilities suggest that the current visual verification methods might soon be obsolete. As generative AI grows more sophisticated, the ease of fabricating convincing ID images and bypassing liveness checks poses a serious challenge to financial institutions. Justin Leroux's observation, "When we can no longer trust our eyes to ascertain whether content is genuine we'll rely on applied cryptography," underscores the urgency for more secure authentication methods.

In response to these developments, private key cryptography and decentralized ID solutions are gaining traction as potential alternatives to traditional KYC. A stable solution must ensure both the ease of customer onboarding and robust security against AI-assisted impersonation threats. As deepfake technology races ahead, the financial sector's defense mechanisms will need to evolve just as rapidly to maintain trust and integrity. techcrunch

Together with VoiceSpin

🚀 Skyrocket Your Sales with AI-powered Auto Dialer

Supercharge your outbound calling efforts and skyrocket your sales with cutting-edge technology. Connect with more leads, engage in meaningful conversations, and close deals in record time. Key Benefits:

  • Make thousands of outbound calls

  • Connect leads with the best-fitting agents for higher conversions

  • Utilize dynamic lead and agent scoring for a substantial boost in conversion rates

Don't miss out on the opportunity to revolutionize your sales strategy. Elevate your business with VoiceSpin.

Try VoiceSpin Now!

Please support our sponsors!

Mastering AI

Top Tools

  • Distribute.so: Transform your website creation game! No more expensive engineers needed. Design and develop content pages and mini-websites in minutes, not months. (website) (sponsored)

  • Vizologi: AI-powered software that generates business insights, ideas, and plans. (website)

  • Trudo AI: Build Open AI apps with intuitive UI & GPT3 fine-tuning. (website)

Practical AI

Unlearning To Build Great AI Apps

In the fast-moving domain of generative AI, product teams are recognizing the need to shift gears. Traditional backtesting, once the bedrock of AI product success, is giving way to new strategies that prioritize user engagement and feedback loops. As outlined by a seasoned head of product at Tome, the most effective teams simultaneously tackle user problems while leveraging the latest AI advancements. They're flipping the classical approach on its head, starting with prototyping to align technological capabilities with user needs, rather than letting user issues alone dictate the development trajectory.

The transition is also reshaping how we gather feedback. Instead of relying on direct user interaction data, generative AI requires a more nuanced approach, integrating feedback mechanisms into the application itself. For example, at Tome, the act of sharing a presentation is used as an indicator of content quality, guiding further enhancement of AI outputs. This pivot from traditional methods of system improvement to a model that captures feedback in a more organic, less intrusive manner is critical for generative AI's evolution. towardsdatascience

Around The World

More News

  • CES 2024 Unveils a Spectacle of Innovation with Live Updates on Breakthrough Gadgets and In-Depth Photo Coverage from the Forefront of Technology's Grand Showcase.

  • Judges in England and Wales Integrate Artificial Intelligence Tools to Streamline the Drafting of Legal Opinions.

  • AI Firms Tackle Misinformation Challenge by Developing Strategies to Mitigate Large Language Model Hallucinations in Effort to Enhance Reliability and Accuracy of Generated Content.

Trending Today

Research Review

Multiple AI Models Help Robots Execute Complex Plans More Transparently

MIT's Improbable AI Lab has innovated a multimodal framework, HiP, to enable robots to perform complex, multi-step tasks more transparently. By utilizing three distinct foundation models, HiP crafts detailed plans drawing on linguistic, physical, and environmental intelligence without the need for paired data sets. This trio of models—comprising a language reasoner, a visual world model, and an action planner—allows robots to adapt to changes in real-time, making the decision-making process both transparent and tractable.

In testing, HiP surpassed comparable systems in tasks requiring adaptability, such as color-matching in block stacking and selective object arrangement. The system's hierarchical structure, with an LLM at its base, enables iterative refinement of plans, akin to the way an author revises an article. Although HiP's potential is currently bounded by the quality of video foundation models, its economical training and use of minimal data showcase the promise of leveraging existing models for complex robotic planning. wired

Written by Isaac R. Ward, Casey Clifton, and Alex Brogan.

Send us feedback at [email protected] and help us provide the best coverage of Artificial Intelligence possible.

Interested in reaching smart business leaders like you? To become an Intelligence Age partner, apply here.

Join the conversation

or to participate.