IBL News

Author: IBL News

  • Facebook Will Launch AI Chatbots With Different Personalities

    Facebook Will Launch AI Chatbots With Different Personalities

    IBL News | New York

    Meta plans to launch a range of AI chatbots that exhibit different personalities and offer recommendations on Facebook in September.

    An example is a character that emulates Abraham Lincoln. Facebook, which seeks to retain and attract users in its battle with social media upstart TikTok, sees these characters as a fun product for people to play with.

    On top of boosting engagement, Facebook’s chatbots can collect vast amounts of data on users’ interests to attract advertisers. Currently, most of Meta’s $117bn a year in revenues comes from advertising.

    Rival companies, such as Andreessen Horowitz-backed start-up Character.ai, have already launched chatbots that feature personalities who generate conversation in the style of individuals, such as Elon Musk and the Nintendo character Mario.

    During an earnings call this month, Zuckerberg told analysts that the company is also building AI agents that can help businesses with customer service and productivity assistants for staff.

    Meta has been investing in generative AI, technology that can create text, images, and code. This month, it released a commercial version of a large language model that could power its chatbots, called Llama 2.
    .

  • NASA and IBM Teamed Up to Build an AI, Open Source Model for Earth Observations

    NASA and IBM Teamed Up to Build an AI, Open Source Model for Earth Observations

    IBL News | New York

    NASA, IBM, and Hugging Face teamed up to create an AI open-source geospatial foundation model for earth observations.

    The project will serve as the basis for innovations in addressing critical environmental challenges since the tool can track deforestation, predict crop yields, and rack greenhouse gas emissions.

    It uses large-scale satellite and remote sensing data, including the Harmonized Landsat and Sentinel-2 (HLS) data, and is accessible to open science users, startups, and enterprises on multi-cloud AI platforms like Watsonx.

    “By combining IBM’s foundation model efforts aimed at creating flexible, reusable AI systems with NASA’s repository of Earth-satellite data, and making it available on the leading open-source AI platform, Hugging Face, we can leverage the power of collaboration to implement faster and more impactful solutions that will improve our planet,” Sriram Raghavan, VP of IBM Research AI, said in a press release.

    NASA estimates that its Earth science missions will generate around a quarter million terabytes of data in 2024 alone.
    .

  • AWS Announced Autonomous AI Agents that Add Key Functionality to Foundation Models

    AWS Announced Autonomous AI Agents that Add Key Functionality to Foundation Models

    IBL News | New York

    Amazon’s AWS cloud provider is showing this month to large customers its service Agents for the Amazon Bedrock foundation model, which lets businesses create chatbots that execute tasks and give more personalized answers drawing from their proprietary data.

    For example, an airline can build a virtual agent that books a flight for a traveler, for instance, based on a customer’s price, destination, and seating requests.

    Another example takes place in the healthcare industry: software vendors can build apps that transcribe and analyze clinical notes after a patient visit. This service is offered now under the name of AWS HeathScribe.

    “Our mission is to make every company an AI company,” said AWS’ Vice President Swami Sivasubramanian during a large AWS event in Manhattan last month.

    Amazon Bedrock is the company’s answer to services announced by Google’s and Microsoft’s cloud AI offering.

    Amazon Bedrock’s API service allows developers to build applications without the need to manage AI infrastructure at all. Because they run on an AWS-managed infrastructure, the models can be scaled on services like EC2 and Lambda and provide low-latency endpoints to enable real-time integration into workflows.

    The two Amazon Bedrock Titan foundation models, Titan Text and Titan Embeddings are pre-trained on large datasets.

    According to Amazon, thousands of customers are now using Amazon Bedrock for various generative AI applications, such as self-service, customer care, text creation, and post-call analysis.
    .

  • ChatGPT Will Feature Prompt Examples and Suggest Replies to Users

    ChatGPT Will Feature Prompt Examples and Suggest Replies to Users

    IBL News | New York

    OpenAI plans to roll out a bunch of updates on ChatGPT this week.

    The chatbot will feature prompt examples at the beginning of a new conversation and will suggest replies to continue the engagement.

    For paid users in the Plus category, ChatGPT will remember the GPT-4 selected model without defaulting back to GPT-3.5.

    Also. for users in the same Plus category, the new Code Interpreter feature will be available, resulting in ChatGPT analyzing data and generating insights across multiple files.

    Finally, OpenAI has implemented keyboard shortcuts: ⌘ (Ctrl) + Shift + C to copy code block and ⌘ (Ctrl) + / to see the complete list.

    On the other hand, OpenAI has launched a web crawler called GPTBot to improve its AI models.

    Users have the possibility to prevent GPTBot from accessing their website’s content and/or allowing certain pages and disallowing others, as explained, as explained on an OpenAI blog post.
    .

    Update August 10, 2023: OpenAI announced that it’s expanding Custom instructions to all users, including those on the free tier of the service.

    These custom instructions allow users to add various preferences and requirements that they want the AI chatbot to consider when responding, so they don’t have to rewrite the same instruction prompts every time they interact with the chatbot.

    For example, a teacher using ChatGPT to create a lesson plan would no longer have to constantly repeat that they teach 3rd grade to receive an appropriately customized response.

  • Zoom Will Use Users’ Data to Train Its AI Models

    Zoom Will Use Users’ Data to Train Its AI Models

    IBL News | New York

    According to recently updated terms of service, Zoom can now train its AI models using users data on product usage.

    The latest update to the video platform’s terms of service, effective as of July 27, on its print, seems to establish Zoom’s right to utilize some aspects of customer data for training and tuning its AI, or machine-learning models. It does not provide an opt-out option.

    These new terms reveal Zoom’s own AI strategy, according to experts.

    Zoom’s terms literally state: “You consent to Zoom’s access, use, collection, creation, modification, distribution, processing, sharing, maintenance, and storage of Service Generated Data for any purpose, to the extent and in the manner permitted under applicable Law, including for the purpose of … machine learning or artificial intelligence (including for the purposes of training and tuning of algorithms and models).”

    “Your content is used solely to improve the performance and accuracy of these AI services,” Zoom wrote in a blog post.

    The update comes amid growing public debate on the extent to which AI services should be trained on individuals’ data. Chatbots such as ChatGPT, Google’s Bard, and Microsoft’s Bing, along with image-generation tools like Midjourney and Stable Diffusion, are trained on swaths of internet text or images.

    Across the generative AI sector, lawsuits have popped up in recent months from authors or artists who say they see their own work reflected in AI tools’ outputs.

    In June, Zoom introduced two new generative AI features, such as a meeting summary tool and a tool for composing chat messages, on a free trial basis for customers. When a user does enable these features and therefore gives his content, Zoom has them sign a consent form allowing Zoom to train its AI models.
    .

     

  • Coursera Issues a Prompt Engineering Class While It Prepares Its AI Toolset

    Coursera Issues a Prompt Engineering Class While It Prepares Its AI Toolset

    IBL News | New York

    Coursera launched several new short-form content, content, courses, and credentials in the last weeks, covering generative AI. This offer includes two timely courses:

    Coursera also shared that it continues developing new generative AI-powered tools and features which are targeted to launch later this year. Among them:

    • Coursera Virtual Coach. It will answer questions, share personalized feedback, and provide quick video lecture summaries and resources, such as a recommended Clip. It will also communicate in different languages and for any education level.

    • Machine Learning Powered Translations. Course readings, lecture video subtitles, quizzes, assessments, peer review instructions, and discussion prompts across 2,000+ courses will be translated from English into Spanish, Arabic, Brazilian Portuguese, French, German, Indonesian, and Thai.
    • AI-Assisted Course Builder. This tool will auto-generate course content, such as course structure, readings, assignments, and glossaries. Authors will be able to integrate their privately-authored content.

     

    https://youtu.be/2NoNdedtUnc

  • Jupyter Releases Generative AI Features Allowing to Generate Notebooks from a Prompt

    Jupyter Releases Generative AI Features Allowing to Generate Notebooks from a Prompt

    IBL News | New York

    The Jupyter organization announced yesterday an open-source project to bring advanced generative AI features to Jupyter Notebook and JupyterLab.

    AWS Jupyter AI offers:

    • A native chat UI conversational assistant in JupyterLab.
    • An %%ai magic that turns the Jupyter Notebook into a reproducible generative AI playground. This works anywhere the IPython kernel runs (JupyterLab, Jupyter Notebook, Google Colab, VSCode, etc.).
    • Support for a wide range of generative model providers and models (AI21, Anthropic, Cohere, Hugging Face, OpenAI, SageMaker, etc.).

     

  • Meta Open-Sourced AudioCraft, a Generative AI Software for Sound and Music

    Meta Open-Sourced AudioCraft, a Generative AI Software for Sound and Music

    IBL News | New York

    Meta this week shared details on AudioCraft, a single open-source code base for music, sound, compression & generation of high-quality, realistic audio from text.

    AudioCraft consists of three models: MusicGen, AudioGen, and EnCodec.

    • MusicGen, trained with Meta-owned and specifically licensed music, generates music from text-based user inputs. [See Demo]

    • AudioGen, which was trained on public sound effects, generates audio from text-based user inputs.

    • EnCodec decoder allows for higher quality music generation with fewer artifacts.

    Meta’s pre-trained AudioGen model lets users generate environmental sounds and sound effects like a dog barking, cars honking, or footsteps on a wooden floor.

    The models are available for research purposes and to further people’s understanding of the technology.

    “We’re excited to give researchers and practitioners access so they can train their own models with their own datasets for the first time and help advance the state of the art,” said Meta in a blog post.

    • Samples of generated audio.

  • “Chatbots are a terrible interface for LLMs,” Expert Designers Say

    “Chatbots are a terrible interface for LLMs,” Expert Designers Say

    IBL News | New York

    Chatbot interfaces look the same as a Google search box, a login form, and a credit card field. “They are not the future; pretty soon, we’re going to get sick of typing all the time, static text is dead,” wrote design expert Amelia Wattenberger.

    Dealing with text prompts, she suggested adding controls to better tailor the responses, as shown below.

     

    .

    • “Even if we’re determined to stick with a chat interface, we can make things easier for users. Recently, my team shipped a prototype named Copilot for Docs, exploring ways to make technical documentation easier for developers to use.” [See it below]

    • “I believe the real game changers are going to have very little to do with plain content generation. Let’s build tools that offer suggestions to help us gain clarity in our thinking.”
    .
    • “Chatbots are a terrible interface for LLMs. Or, at the very least, we can add controls, information, and affordances to our chatbot interfaces to make them more usable. I can’t wait to see the field become more mature and for us to start building AI tools that embrace our human abilities.”

     

    • GitHub: Awesome ChatGPT Prompts 

     

     

  • Ukrainian Startup Preply Raised $70M for Its Language Learning App

    Ukrainian Startup Preply Raised $70M for Its Language Learning App

    IBL News | New York

    The live language learning app Preply, known for its large number of human tutors (currently, 35,000 in 50 languages), raised $70 million in equity and debt to extend its series C to $120 million.

    The round was led by Horizon Capital, Hoxton Ventures, owl Ventures, and other investors.

    Preply, which made its name out of its network of tutors, and differentiated itself from the likes of Doulingo’s automatic approach, said in an announcement that it plans now to use the funding to double down on AI to scale its service.

    Founded in Ukraine, the company has already incorporated an AI teaching assistant that allows users to create exercises, grammar explanations, conversation starters, instructional planning, and after-lesson support. It’s also been building AI analytics to help understand how students are doing.

    Preply has Datadog, GroupM, and Bain among its customers.
    .