top of page
  • Writer's pictureSTR Nexus

AI: Tech Giants Invest Billions in Machine Learning Tools

Staying up to date with the rapidly evolving AI industry can be challenging. Until an AI can handle that task, here's a convenient summary of last week's developments in the realm of machine learning, including noteworthy research and experiments that may have gone unnoticed.

AI: Tech Giants Invest Billions in Machine Learning Tools

The competitive landscape in AI, particularly in the generative AI subfield, is heating up. Dropbox has launched its corporate venture fund, Dropbox Ventures, with a focus on supporting startups developing AI-powered products that shape the future of work. AWS has also entered the scene with a $100 million program to fund generative AI initiatives led by its partners and customers.

Money is pouring into the AI space from various sources. Salesforce Ventures plans to invest $500 million in startups working on generative AI technologies. Workday has added an additional $250 million to its existing VC fund, specifically targeting AI and machine learning startups. Accenture and PwC have also announced substantial investments of $3 billion and $1 billion, respectively, in AI.

However, it is important to question whether money alone can solve the outstanding challenges in the AI field. During a panel discussion at a recent Bloomberg conference in San Francisco, Meredith Whittaker, president of the secure messaging app Signal, highlighted the increasing opacity of the underlying technology behind popular AI applications. She gave an example of a person being denied a loan without their knowledge that a system powered by a Microsoft API, utilizing scraped social media data, determined their lack of creditworthiness. Whittaker emphasized the need to address the existing power hierarchy rather than solely relying on capital.

Structural change poses a greater challenge than merely seeking funding, especially when it may not favor those in power. Whittaker warned of the consequences if sufficient resistance is not put up against the entrenchment and naturalization of power under the guise of intelligence, leading to extensive surveillance and limited agency over individual and collective lives.

While financial investments and advancements in AI continue, other noteworthy headlines in the past few days include:

  • DeepMind's RoboCat: DeepMind has developed a versatile AI model called RoboCat, capable of performing various tasks across different types of robotic arms.

  • VRB (Vision-Robotics Bridge): CMU Robotics Institute showcased VRB, an AI system designed to train robotic systems by observing human actions and executing similar tasks.

  • Otter's AI-powered chatbot: Automatic transcription service Otter introduced an AI chatbot that enables participants to ask questions during and after a meeting, fostering collaboration with teammates.

  • EU's call for AI regulation: The European Consumer Organisation (BEUC) urged urgent investigations into the risks associated with generative AI, emphasizing the need for timely AI regulation.

  • AI-powered features by Vimeo: Vimeo launched a suite of AI-powered tools that assist users in script creation, teleprompter-based recording, and removing disfluencies from recordings.

  • Funding for synthetic voices: ElevenLabs, a platform for creating synthetic voices, secured $19 million in a recent funding round, although some concerns have arisen due to misuse of the technology.

  • Audio-to-text transcription by Gladia: French AI startup Gladia introduced a platform utilizing OpenAI's Whisper transcription model to convert audio into near real-time text via an API, offering affordable transcription services.

  • Generative AI integration by Harness: Harness, a startup providing a toolkit for developers, integrated AI into its platform, enabling automatic resolution of build and deployment failures, identification and remediation of security vulnerabilities, and cost optimization for cloud services.

In other machine learning news:

  • Yejin Choi's keynote: Yejin Choi, a UW professor and MacArthur Genius grant recipient, addressed the limitations of today's powerful language models like GPT-4, emphasizing the importance of understanding their boundaries.

  • Rod Brooks' historical perspective: Brooks provided a historical overview of core machine learning concepts, highlighting their continued relevance and the field's reliance on past research.

  • CVPR conference highlights: Notable papers and presentations from the CVPR conference included VISPROG's complex visual manipulation capabilities, a unified approach to autonomous driving by a Chinese research group, and DynIBaR's advancements in interacting with videos using dynamic Neural Radiance Fields.

  • DreamBooth for deepfakes: DreamBooth, a project focused on creating deepfakes, showcases advancements in image manipulation techniques, which have both positive and potentially negative consequences.

  • Student paper award: The best student paper award recognized a method for comparing and matching meshes or 3D point clouds, crucial for real-world perception applications.

Additionally, Intel's LDM3D model for generating 3D 360 imagery and Meta's voice synthesis tool called Voicebox, capable of replicating voices with minimal clean data, were notable highlights.

These developments provide a glimpse into the evolving landscape of AI and machine learning, with both opportunities and challenges ahead.

2 views0 comments


Commenting has been turned off.
bottom of page