AI News Update | Week 23-30 Sep 23

Meta, formerly known as Facebook, is making significant strides in the realm of artificial intelligence and augmented reality. Recent announcements from the company highlight a series of AI-driven tools, experiences, and innovative hardware, emphasizing Meta’s commitment to enhancing user interactions and creativity across its platforms.

Features:

1. Meta AI Assistant: A conversational AI assistant designed to provide real-time information across WhatsApp, Messenger, and Instagram. Soon, it will also be integrated into Ray-Ban Meta smart glasses and Quest 3, offering users a seamless AI experience across devices.
2. Facebook-Streaming Glasses: In collaboration with Ray-Ban, Meta has introduced smart glasses capable of streaming content directly from Facebook, marking a significant step in AR integration with social media.
3. AI Characters: A diverse roster of 28 AI personalities, including representations of celebrities, is set to enhance user interactions on various platforms.
4. AI-Powered Creative Tools: Including customizable AI stickers and AI-driven image editing tools like “Restyle” and “Backdrop.”
5. Partnership with Microsoft’s Bing: This collaboration aims to provide real-time web results, setting Meta AI apart from other AI models.
6. Generative AI: The ability of the Meta AI to generate images, similar to OpenAI’s DALL-E, offering high-resolution photos based on user prompts.

Benefits:

1. Enhanced User Experience: The introduction of AI characters, the Meta AI Assistant, and Facebook-streaming glasses promises more engaging and dynamic interactions for users.
2. Creative Empowerment: AI-driven tools offer users the ability to co-create and reimagine images, fostering creativity.
3. Seamless Integration: With the AI Assistant embedded within popular chat applications and devices, users can access real-time information without leaving their conversation context.
4. Safety Measures: Meta has implemented safeguards in its AI to prevent the dissemination of misleading or harmful information.

Other Noteworthy Points:

• Meta’s vision emphasizes transparency, marking AI-generated content to distinguish it from human-created content.
• The company’s commitment to responsible AI development is evident, with a focus on user safety and continuous improvements based on feedback.
• Meta’s vast user base across its messaging apps is seen as a competitive advantage, positioning the company to potentially lead the AI chatbot race against competitors like OpenAI’s ChatGPT.
• Future integrations may leverage public user data from Instagram and Facebook to enhance the AI’s capabilities, offering unique experiences that other chatbots might not be able to replicate.

OpenAI, renowned for its advancements in artificial intelligence, has unveiled a significant upgrade to ChatGPT. This enhancement equips ChatGPT with the ability to see, hear, and speak, marking a transformative step in user interaction with AI.

Features:

1. Voice Interaction: Users can now engage in voice conversations with ChatGPT. This feature is powered by a new text-to-speech model, capable of generating human-like audio. The voice capability offers five distinct voices, crafted in collaboration with professional voice actors.
2. Image Interaction: ChatGPT can now analyze and discuss images shared by users. Whether it’s troubleshooting a device, planning a meal based on fridge contents, or analyzing complex work-related graphs, ChatGPT offers insights using its image understanding capabilities.
3. Drawing Tool: Available on the mobile app, this tool allows users to focus on specific parts of an image, guiding ChatGPT’s attention to particular areas.

Benefits:

1. Enhanced User Engagement: The addition of voice and image capabilities offers a more intuitive and dynamic interaction experience for users.
2. Versatility: From casual conversations about landmarks to assisting with academic or professional tasks, ChatGPT’s expanded capabilities cater to a wide range of user needs.
3. Accessibility: Collaborations with platforms like “Be My Eyes” highlight OpenAI’s commitment to making AI tools accessible and beneficial for visually impaired individuals.

Other Noteworthy Points:

• OpenAI emphasizes a gradual rollout strategy for its tools, allowing for continuous refinement and risk mitigation over time.
• The new voice technology presents both opportunities and challenges. While it can be used for creative and accessibility-focused applications, there are concerns about potential misuse, such as impersonation or fraud.
• Vision-based models, while powerful, come with their own set of challenges. OpenAI has taken measures to ensure that ChatGPT respects individuals’ privacy and provides accurate image interpretations.
• OpenAI is transparent about ChatGPT’s limitations, especially in specialized topics and non-English text transcription.
• Access to these new features will be expanded to Plus and Enterprise users in the coming weeks, with plans to roll out to other user groups soon after.

1. Meta Platforms has trained its new Meta AI virtual assistant using public posts from Facebook and Instagram, deliberately excluding private posts and chats to prioritize user privacy, according to Nick Clegg, Meta’s President of Global Affairs.

Speaking at the company’s annual Connect conference, Clegg emphasized that datasets with significant personal information were avoided, and public data constituted the “vast majority” of training material. He highlighted that content from platforms like LinkedIn was intentionally not used due to privacy reasons. This approach comes amid scrutiny of tech giants like Meta, OpenAI, and Google for using web data without consent to train AI models. These companies are currently evaluating the handling of private or copyrighted content that their AI might replicate, amidst legal challenges over copyright breaches.

Meta AI stood out as a key product introduced by CEO Mark Zuckerberg at the recent Connect conference, which this year was predominantly centred on artificial intelligence, a shift from previous years’ emphasis on augmented and virtual reality.

2. Jony Ive, Apple’s former chief design officer, and Sam Altman, CEO of OpenAI, are reportedly collaborating on a groundbreaking artificial intelligence “hardware device.”

This initiative is backed by discussions with SoftBank’s Masayoshi Son, with potential funding exceeding $1 billion, as indicated by the Financial Times. The specifics of the device, often referred to as the “iPhone of artificial intelligence,” remain undisclosed. However, sources cited by The Information suggest the duo is exploring what “new hardware for the AI age could look like.” Altman’s history includes partnerships with ex-Apple designers, such as Thomas Meyerhoffer, who designed the retina-scanning Orb for Altman’s Worldcoin project. Additionally, Altman has ties with Humane, a startup developing an AI wearable device, and has previously collaborated with ex-Apple employees. SoftBank’s CEO, Masayoshi Son, a significant investor in AI startups, is also actively involved in these discussions, emphasizing his vision to position SoftBank as a pivotal player in the AI revolution.

2. AI-driven market intelligence platform, AlphaSense, has achieved a significant milestone in its growth trajectory.

The company recently completed a funding round, raising an impressive $150 million. This influx of capital has propelled AlphaSense’s valuation to a noteworthy $2.5 billion. The successful funding round and the subsequent valuation underscore the robust demand in the market for innovative artificial intelligence (AI) startups. As the AI sector continues to evolve, companies like AlphaSense are emerging as frontrunners, capturing the attention of investors and industry stakeholders alike.

2. OpenAI, backed by tech giant Microsoft, has unveiled a pivotal enhancement to its widely recognized chatbot, ChatGPT. In a recent announcement, OpenAI revealed that ChatGPT users would now possess the capability to browse the internet directly through the platform.

This significant update extends the chatbot’s data access horizon beyond its previously established limit in September 2021.

The integration of browsing capabilities is anticipated to revolutionize the way users interact with ChatGPT. By granting the chatbot access to real-time internet data, users can expect more comprehensive, up-to-date, and contextually relevant responses. This move not only amplifies the utility of ChatGPT but also positions it as a frontrunner in the realm of AI-driven chat interfaces, bridging the gap between static information and dynamic web content.

As AI chatbots continue to permeate various sectors, from customer service to personal assistants, the ability to access and relay current information from the web further solidifies ChatGPT’s standing in the market. OpenAI’s commitment to continuous improvement and user-centric enhancements is evident in this update, promising a richer and more informed conversational experience for all ChatGPT users.

5. Google’s AI chatbot, Bard, has expanded its capabilities beyond mere web searches. The tech giant has integrated Bard with Gmail, Docs, and Drive, enabling users to seamlessly retrieve and summarize content from these platforms.

This enhancement aims to streamline information retrieval, eliminating the need for users to manually sift through vast amounts of data. For instance, Bard can highlight key points from a document stored in Drive or summarize the contents of an email.

While these integrations, termed as “extensions” by Google, promise efficiency, they also raise potential privacy concerns. Addressing this, Google assures that personal data accessed by Bard will neither be used to train the chatbot’s public model nor be viewed by human reviewers. Users are required to opt-in for these integrations and have the freedom to disable them at any given time.

Beyond Gmail, Docs, and Drive, Bard’s extensions also encompass Maps, YouTube, and Google Flights. This allows users to extract real-time flight details, discover nearby attractions, or even fetch specific YouTube videos. Jack Krawczyk, Bard’s product lead, emphasized the significance of these integrations, noting it’s the first instance of a language model product integrating with personal data. Google’s vision for Bard includes further integrations across its product suite and potential collaborations with external partners.

Additional improvements to Bard include the “Google It” button, which allows users to verify Bard’s responses against Google Search results. This feature highlights corroborated information in green and unvalidated answers in orange, offering users clarity on the accuracy of Bard’s responses. Since Bard’s introduction in February, Google has consistently enhanced its features, recently integrating it with Google Lens and adding functionalities for Google Sheets.

6. OpenAI’s Valuation Skyrockets Amid Share Sale Talks

OpenAI, the AI powerhouse behind ChatGPT, is in advanced discussions with investors about a potential share sale, which could elevate its valuation to between $80 billion and $90 billion. This marks a significant jump from earlier valuations this year.

Microsoft, holding a 49% stake in OpenAI, is set to benefit considerably from this valuation surge. OpenAI’s revenue projections are robust, with an anticipated $1 billion for the current year and expectations to hit multibillion-dollar revenues by 2024. The primary sources of these revenues are the premium versions of ChatGPT and licensing its AI technology to businesses.

The proposed share sale is structured to allow OpenAI employees to sell their shares, emphasizing the company’s commitment to its workforce. Since ChatGPT’s launch last November, OpenAI has solidified its position in the AI sector, even as tech giants like Google and Meta venture deeper into the AI domain.

Historical investors in OpenAI include Sequoia Capital and Khosla Ventures, but Microsoft remains its primary external funder. CEO Sam Altman has attracted significant investor interest, including from Softbank. Altman has been a vocal advocate for AI regulation and safety standards and has clarified that OpenAI has no intentions of going public or being acquired.

OpenAI aims to ensure Microsoft remains a minority stakeholder, emphasizing its vision of maintaining independence in the rapidly evolving AI landscape.

7. Microsoft is testing a DALL-E-powered feature in Microsoft Paint for Windows 11 named “Paint Cocreator.”

This tool generates artwork from user-provided descriptions, offering variations for users to refine. While the latest DALL-E model’s accuracy in converting text to images is under observation, Microsoft has implemented measures to prevent generating inappropriate images.

The Paint Cocreator is set for a wider release soon but is currently available to a select group of Windows Insiders. Users joining the preview are granted 50 credits for creating images, hinting at potential future charges.

This initiative is part of Microsoft’s push to integrate AI-driven updates into Windows 11, including enhancements in the Photos app and the introduction of the Windows Copilot sidebar, an AI assistant.

1. In a significant development within the entertainment industry, Hollywood studios are poised to retain the right to utilize writers’ work for training artificial intelligence models.

This revelation comes from a tentative labour agreement between the studios and the writers, as reported by the Wall Street Journal. While the exact terms and conditions of this agreement have not been disclosed, it underscores the growing integration of AI technology within the realm of creative content production.

The move is indicative of the broader trend of industries leveraging AI to enhance their operations and offerings. For the Hollywood studios, the ability to train AI models on writers’ work could potentially lead to innovations in content creation, script analysis, and other areas. However, it also raises questions about intellectual property rights, compensation, and the potential impact on the creative process.

The agreement, once finalized, could set a precedent for how other sectors approach the intersection of AI and creative content. As AI continues to permeate various facets of our lives, its integration into the arts and entertainment sector is a testament to its transformative potential.

2. Tencent-backed AI Chip Startup Secures Significant Investment

In a significant stride for China’s AI industry, Enflame, an artificial intelligence (AI) chip startup backed by the tech giant Tencent, has announced a substantial fundraising achievement. The company has successfully procured 2 billion yuan (approximately $273.68 million) in its recent funding round. This financial boost is particularly noteworthy given the participation of funds associated with a government authority based in Shanghai.

This investment underscores the escalating interest in the AI chip sector, a domain that is rapidly gaining traction in the global tech arena. Enflame’s success in securing such a sizable investment also highlights the confidence that both private and state-linked entities have in the potential and future of AI-driven technologies.

The involvement of Tencent, a dominant player in the tech industry, further elevates the significance of this funding round. As AI continues to permeate various sectors, from healthcare to entertainment, the demand for specialized AI chips is expected to surge. Enflame’s recent financial backing positions it well to capitalize on this growing demand and reinforces its standing in the competitive AI chip market.

3. Luda Gains $7M Funding for AI Training Simulation.

Luda, specializing in user-trained AI agents, has introduced its Real-Time Reinforcement Learning (RT-RL) system. This system allows users to create AI entities that interact based on physics without manual coding. To bolster its AI initiatives, Luda secured $7 million in funding.

Bitkraft Ventures and Compound led the funding, with participation from notable figures like Jeff Dean, Google’s chief scientist. Scott Rupp of Bitkraft highlighted Luda’s potential to redefine user-AI interaction, especially in gaming.

Luda’s platform, Mels, animates AI agents using generative AI, optimized for user-generated content gameplay. CEO Vijay Sundaram envisions Luda’s technology as a bridge between human creativity and interactive AI agents.

4. Slope’s AI-Driven B2B Payments Platform Secures $30M

San Francisco-based startup, Slope, is transforming B2B payments using AI. The platform, which integrates proprietary technology with OpenAI’s GPT-3.5 Turbo, addresses challenges like delayed payments and fraud. The company’s recent funding round, led by Union Square Ventures and featuring contributions from OpenAI CEO Sam Altman, brought in $30 million, pushing Slope’s total funding to $187 million.

CEO Lawrence Lin Murata emphasized the platform’s potential in the B2B sector, while Chief Product Officer Alice Deng highlighted its comprehensive tools, from online payments to risk assessment. A standout feature, Slope Timeline, offers real-time payment and shipping updates. Slope’s AI capabilities, especially SlopeGPT, are pivotal in assessing buyer creditworthiness and detecting fraud.

5. Mistral AI Launches High-Efficiency Mistral 7B Model

Paris-based startup, Mistral AI, has introduced its large language AI model, Mistral 7B. With 7.3 billion parameters, this model is reported to outperform larger counterparts, including Meta’s Llama 2 13B.

Mistral 7B, open-sourced under the Apache 2.0 license, is adept at English tasks and coding capabilities, making it a prime choice for enterprise applications. Benchmarks reveal its superior performance over competitors like Llama 2 7B and 13B. While it showcases commendable coding task results, it slightly lags behind the finetuned CodeLlama 7B in specific metrics.

Mistral AI has future plans for an advanced 2024 model with enhanced reasoning and multilingual features. Currently, Mistral 7B offers versatile deployment options across various platforms.

6. Meta’s Llama Ecosystem: Embracing Open Source and Community Collaboration

Meta has emphasized its commitment to the open-source community with the introduction of Llama 2, aiming to make state-of-the-art AI technology accessible to all. The company believes that open collaboration fosters safety and better alignment in AI development. Drawing parallels with PyTorch, Meta highlights the value of open-source initiatives in three main areas:

1. Research: Open collaboration accelerates the incorporation of new techniques, performance optimizations, tools, and safety measures.
2. Enterprise and Commercialization: By allowing enterprises and startups to build on Meta’s technology, the company gains insights into use cases, safe model deployment, and potential opportunities.

• Developer Ecosystem: Large Language Models (LLMs) have transformed AI development. An open community approach enables rapid adoption of emerging tools and methodologies, enhancing Meta’s internal capabilities.

Meta’s vision for the future of AI encompasses:

• Multimodal Experiences: Beyond text, AI should embrace various modalities for a more immersive generative experience.
• Safety and Responsibility: With generative AI reshaping responsible AI, Meta aims to emphasize safety, develop new tools, and foster partnerships.
• Community Focus: Meta envisions a developer community with a voice, agency, and innovative capabilities.

During the Meta Connect event, the spotlight was on Llama models and the future of open access. Meta encourages individuals to explore Llama 2, attend Connect Sessions, and delve into resources available on their website.

7. Anthropic Secures $4 Billion Investment from Amazon for AI Development.

Anthropic, a leading AI firm, has forged a partnership with Amazon, marked by a significant investment of up to $4 billion. This collaboration aims to merge Anthropic’s pioneering safety research with the expansive infrastructure of Amazon Web Services (AWS). The goal is to make advanced, safe, and steerable AI more accessible to AWS users.

Under this agreement, AWS will serve as Anthropic’s chief cloud provider. This grants Anthropic access to AWS’s state-of-the-art Trainium and Inferentia chips, enhancing AI model training and deployment capabilities. Additionally, there will be expanded support for Amazon Bedrock, facilitating secure model customization.

Anthropic’s flagship model, Claude 2, will be available for diverse tasks, from dialogue generation to intricate reasoning, boasting a 100,000 token context window for processing vast domain-specific data. Notably, enterprises like LexisNexis and Bridgewater Associates have integrated Anthropic models through Amazon Bedrock.

Both companies are deeply committed to responsible AI development, actively participating in global safety initiatives. With this collaboration, Amazon will acquire a minority stake in Anthropic, ensuring the AI firm has the resources for continued innovation in AI safety and research.

Bogdan Iancu is a seasoned entrepreneur and strategic leader with over 25 years of experience in diverse industrial and commercial fields. His passion for AI, Machine Learning, and Generative AI is underpinned by a deep understanding of advanced calculus, enabling him to leverage these technologies to drive innovation and growth. As a Non-Executive Director, Bogdan brings a wealth of experience and a unique perspective to the boardroom, contributing to robust strategic decisions. With a proven track record of assisting clients worldwide, Bogdan is committed to harnessing the power of AI to transform businesses and create sustainable growth in the digital age.