IBL News

Author: IBL News

  • NVIDIA’s New Open-Source “NeMo Guardrails” Prevents AI Chatbots From Hallucinating

    NVIDIA’s New Open-Source “NeMo Guardrails” Prevents AI Chatbots From Hallucinating

    IBL News | New York

    NVIDIA released today an open-source software called NeMo Guardrails that can prevent AI chatbots from “hallucinating” wrong facts — such as saying incorrect facts, talking about harmful subjects, or opening up security holes.

    It’s a layer of software that sits between the user and the LLM (Large Language Model) or other AI tools. It heads off bad outcomes or bad prompts before the model spits them out.

    The “hallucination” issue with the latest generation of large language models is currently a major blocking point for businesses.

    The company designed this software to work with all LLM-based conversational applications, including OpenAI’s ChatGPT and Google’s Bard.

    NeMo Guardrails enables developers to define user interactions and integrate these guardrails into any application using a Python library.

    It can run on top of LangChain, an open-source toolkit used to plug third-party applications into LLMs.

    In addition, it works with Zapier and a broad range of LLM-enabled applications.

    Developers can create new rules quickly with a few lines of code by setting three kinds of boundaries:

    • Topical guardrails prevent apps from veering off into undesired areas. For example, they keep customer service assistants from answering questions about the weather.
    • Safety guardrails ensure apps respond with accurate, appropriate information. They can filter out unwanted language and enforce that references are made only to credible sources.
    • Security guardrails restrict apps from making connections only to external third-party applications known to be safe.

    “You can write a script that says, if someone talks about this topic, no matter what, respond this way,” said Jonathan Cohen, Vice President of Applied Research at NVIDIA.

    • “You don’t have to trust that a language model will follow a prompt or follow your instructions. It’s actually hard coded in the execution logic of the guardrail system what will happen.”

    • “If you have a customer service chatbot, designed to talk about your products, you probably don’t want it to answer questions about our competitors.”

    • “You want to monitor the conversation. And if that happens, you steer the conversation back to the topics you prefer.”

    The company said, “Much of the NeMo framework is already available as open-source code on GitHub.  Enterprises also can get it as a complete and supported package, part of the NVIDIA AI Enterprise software platform. NeMo is also available as a service. It’s part of NVIDIA AI Foundations, a family of cloud services for businesses that want to create and run custom generative AI models based on their own datasets and domain knowledge.”

    NVIDIA, which tries to maintain its lead in the market for AI chips by simultaneously developing software for machine learning, mentioned the case of South Korea’s leading mobile operator who built an intelligent assistant that’s had 8 million conversations with its customers, as well as a research team in Sweden that employed NeMo to create LLMs that can automate text functions for the country’s hospitals, government and business offices.

    Other AI companies, including Google and OpenAI, have used a method called reinforcement learning from human feedback to prevent harmful outputs from LLM applications. This method uses human testers who create data about which answers are acceptable or not and then train the AI model using that data.

    • NVIDIA Developer Blog: NVIDIA Enables Trustworthy, Safe, and Secure Large Language Model Conversational Systems


    [Disclosure: IBL Education, the parent company of IBL News, works for NVIDIA on projects related to Online Education]

  • Snapchat Makes its Chatbot ‘My AI’ Free but It Gets Negative Reaction

    Snapchat Makes its Chatbot ‘My AI’ Free but It Gets Negative Reaction

    IBL News | New York

    Snapchat made its OpenAI’s powered bot, named ‘My AI’ free to a global audience. Users’ reaction was negative as they couldn’t remove it from the feed unless they were paid subscribers.

    The company upgraded the tool with new functionalities, such as the ability to add it to group chats with friends with an @mention, get recommendations for places on Snap Map and Lenses, and replay Snaps.

    When it launched in February, it was only available to paid subscribers. It also became problematic as it provided some unsafe answers, such as when it suggested how to mask the smell of alcohol and pot at a birthday party.

    The company has since upgraded the tool with new features, including the ability to generate a visual response and upcoming personalization features that will allow users to name My AI and give it more of an identity.

    The feature remains available only to those with a Snapchat+ subscription, which costs $3.99 per month and could be driving upgrades. Snap also announced that Snapchat+ now has more than 3 million subscribers.

  • Google’s Bard Will Generate, Debug, and Provide Explanations for Code

    Google’s Bard Will Generate, Debug, and Provide Explanations for Code

    IBL News | New York

    Google’s Bard upcoming chatbot will emphasize coding tasks, featuring the ability to generate, debug, and provide explanations for code. This has been one of the users’ — especially, code beginners’ — top requests, according to the tech giant, engadget.com explains.

    Bard will write in 20 programming languages, including C++, Java, and Python. It also will export code to Colab, Google’s cloud notebook environment for Python, and write functions for Sheets.

    However, the code might have errors, and Google advises double-checking.

    ChatGPT has the ability to write and improve existing code in several languages as well.

  • US Homeland Security Creates an Artificial Intelligence Task Force

    US Homeland Security Creates an Artificial Intelligence Task Force

    IBL News | New York

    U.S. Homeland Security will create an AI-focused task force to protect critical infrastructure, and combat criminal activity and disinformation.

    Secretary Alejandro Mayorkas [in the picture above] said yesterday that his department “will lead in the responsible use of AI to secure the homeland and in defending against the malicious use of this transformational technology.”

    Mayorkas said the Artificial Intelligence Task Force would also explore how AI could be used to do work like screening cargo coming into the country for illicit goods, like fentanyl or products made with slave labor.

    Mayorkas also urged efforts to use AI to secure electric grids and water supply systems, both of which have been feared to be potential targets of adversaries.

    He said any move to regulate AI would have to find a “sweet spot” where the government could develop guardrails without stifling innovation.

  • Stability Releases Its New LLM, Open-Source, and Free to Use Commercially

    Stability Releases Its New LLM, Open-Source, and Free to Use Commercially

    IBL News | New York

    Stability AI, the startup behind the generative AI art tool Stable Diffusion, released this week an open-source LLM with 15 to 65 billion parameters to follow.

    StableLM can generate text and code and is an alternative to proprietary OpenAI. “It’s a massive deal, that has the potential to change up everything in AI,” said an expert.

    It can be used for commercial or research purposes, under the CC BY-SA-4.0 license.

    This release is available in “alpha” on its GitHub and Hugging Spaces, a platform for hosting AI models and code.

    It builds on the company’s open-sourcing record with EleutherAI, a nonprofit research hub.

    In 2022, it released the image model Stable Diffusion. Earlier releases included GPT-J, GPT-NeoX, and the Pythia suite, which were trained on The Pile open-source dataset — a mix of internet-scraped text samples from websites including PubMed, StackExchange, and Wikipedia.

    These open-source models, including Cerebras-GPT and Dolly-2, might demonstrate that small and efficient models can deliver performance with appropriate training, according to experts.

    StableLM is trained on a new experimental dataset built on The Pile, but three times larger with 1.5 trillion tokens of content.

    “The richness of this dataset gives StableLM surprisingly high performance in conversational and coding tasks, despite its small size of 3 to 7 billion parameters (by comparison, GPT-3 has 175 billion parameters),” said the company in a blog post.

    Stability is also releasing a set of research models that are instruction fine-tuned. They use a combination of five recent, noncommercial, research-only, open-source datasets: Alpaca, GPT4All, Dolly, ShareGPT, and HH.

    “Language models will form the backbone of our digital economy, and we want everyone to have a voice in their design,” said Stability AI.

    In addition, Stability is working on a crowd-sourced RLHF program and an open-source dataset for AI assistants, Open Assistant.

    Stability AI didn’t say whether StableLM hallucinates and includes offensive language and views. “Given that The Pile contains profane, lewd, and otherwise fairly abrasive language, it wouldn’t be surprising if that were the case,” wrote TechCrunch.

    “This is expected to be improved with scale, better data, community feedback, and optimization,” responded Stability.

    Even commercialized models like GPT-4, which have filters and human moderation teams in place, have been shown to spout toxicity.

    Some researchers have criticized the release of open-source models, arguing that they could be used for unsavory purposes like creating phishing emails or aiding malware attacks.

    According to Semafor, Stability AI — which raised over $100 million at a valuation of $1 billion — is under pressure to monetize its efforts. Its CEO Emad Mostaque has hinted at plans to IPO.

    The company is facing legal cases that allege that it infringed on the rights of millions of artists by developing AI art tools using web-scraped, copyrighted images.

  • Open-Source Initiatives Challenge Closed, Proprietary AI Systems With New LLMs

    Open-Source Initiatives Challenge Closed, Proprietary AI Systems With New LLMs

    IBL News | New York

    Several startups, collectives, and academics have released a wave of new large language models (LLMs) as open source, trying to challenge the closed, proprietary AI systems such as OpenAI and Anthropic.

    These private organizations, knowing that state-of-the-art LLMs require huge compute budgets — OpenAI reportedly used 10,000 Nvidia GPUs to train ChatGPT — and deep ML expertise have refused to open up their models. They rely on API distribution instead.

    The data, source code, or deep learning programming, of the model weights, remain hidden from public scrutiny.

    Open-source initiatives state that they are seeking to democratize access to LLMs.

    Two weeks ago, Databricks announced the ChatGPT-type Dolly, which was inspired by Alpaca, another open-source LLM released by Stanford in mid-March.

    Stanford’s Alpaca used the weights from Meta’s LLaMA model that was released in late February.

    LLaMA was hailed for its superior performance over models such as GPT–3, despite having ten times fewer parameters.

    Other open-source LLaMA-inspired models have been released in recent weeks, such as:

    – Vicuna, a fine-tuned version of LLaMA that apparently matches GPT-4 performance;

    – Koala, a model from Berkeley AI Research Institute;

    – ColossalChat, a ChatGPT-type model part of the Colossal-AI project from UC Berkeley.

    Some of these open-source models have even been optimized to run on the lowest-powered devices, from a MacBook Pro down to a Raspberry Pi and an old iPhone.

    However, none of these open-source LLMs are available yet for commercial use, as the LLaMA model is not released for commercial use.

    In addition, the OpenAI GPT-3.5 terms of use prohibit using the model to develop AI models that compete with OpenAI.

    In March, the free-software community Mozilla announced an open-source initiative for developing AI, saying they “intend to create a decentralized AI community that can serve as a ‘counterweight’ against the large profit-focused companies.”

    .

  • Elon Musk Creates His Own Generative AI Company Joining the Race for the LLMs

    Elon Musk Creates His Own Generative AI Company Joining the Race for the LLMs

    IBL News | New York

    Elon Musk has created through his family office a new company named X.AI, intended to compete with OpenAI, Anthropic, Adept, and StabilityAI, and other startups in the artificial intelligence generative space.

    The billionaire entrepreneur owner of Tesla, Space X, and Twitter now plans to attract investors while seeking to recruit a team of top AI researchers and engineers.

    According to Financial Times, for the new project, Musk has secured thousands of high-powered Nvidia’s GPU processors, the high-end chips required for building a large language model.

    Elon Musk is, for now, the company’s only director. The private firm, incorporated in Nevada on March 9, lists as secretary Jared Birchall, the ex-Morgan Stanley banker who manages Musk’s wealth.

    Musk left the board of OpenAI in 2018 amid clashes with its management, including over attitudes to AI safety.

    Since then, Musk became increasingly vocal in his fears of broader existential threats from AI systems.

    He also criticized OpenAI for becoming, in his view, less transparent and too focused on commercialization in its pursuit of advanced AI.

    Musk could use Twitter content as data to train its language model and tap Tesla for computing resources. Its homegrown supercomputer, Dojo, is used to train Tesla’s Autopilot self-driving system.
    .

     

  • In the Battle of the Chatbots, Who Is More Efficient GPT-4 or Google’s Bard?

    In the Battle of the Chatbots, Who Is More Efficient GPT-4 or Google’s Bard?

    IBL News | New York

    In the battle of the chatbots between Microsoft-backed Open AI’s GPT-4 and Google’s Bard, these conversational engines are trained on websites, books, documents, and Wikipedia, to generate responses to predict the likely next word in a sentence mimicking human speech.

    Bard is trained to engage in natural-sounding dialogue while GPT-4 seeks to generate in-depth replies on a broad range of topics and has knowledge of events until September 2021, according to an analysis by Financial Times, that tested both.

    Both companies have been opaque about how their models were built.

     

  • Generative AI NOLEJ Releases Its Instructional Design Content Creator

    Generative AI NOLEJ Releases Its Instructional Design Content Creator

    IBL News | New York

    French generative AI startup NOLEJ has made its OpenAI-based instructional content generator available for educators as a free trial, with five packages.

    The AI tool can turn video, audio, and text content into micro-learning interactive videos, quizzes, and flashcards. Users can download the package as SCORM or HTML5 code, or copy the associated embed code and paste it into most websites or LMSs.

    The company reported successful testing of its GPT-3.5 version with 2,500 educators and announced a strategic collaboration with OpenAI.

    “This collaboration aims to create new tools that will enhance the learning experience even further and open up new possibilities for educators,” said Bodo Hoenen, CEO and co-founder of NOLEJ.

    Other educational companies collaborating with OpenAI include Shutterstock, Duolingo, and Khan Academy.

    Nolej AI was tested by Campus Technology’s sister publication, The Journal, for its report.
    .

     

    https://youtu.be/z7dTwxKE-v0

     

  • Coursera Will Release Its AI Chatbot Over the Coming Months

    Coursera Will Release Its AI Chatbot Over the Coming Months

    IBL News | New York

    Coursera announced this week that new generative AI features will be implemented “over the coming months.”

    “These innovations, intended to build a personalized learner experience, mark a new chapter for teaching and learning on Coursera,” said Shravan Goli, Chief Operating Officer of Coursera in a blog post. [Update: June 19, 2023] 

    • Coursera Coach: Virtual coach that answers questions and shares personalized feedback. It also provides quick video lecture summaries and resources, such as a recommended clip, to help learners better understand a specific concept. Coursera said that a pilot will be launched “in the coming months.”
    • Coursera Author: AI-powered course building features to auto-generate content, such as, overall structure, readings, assignments, and glossaries. It will help educators reduce the time and cost of producing high-quality content. A pilot launch is scheduled for “later this year.”
    • Massive translation from English into Spanish, French, German, Arabic, Portuguese, Thai, and Indonesian of 2,000+ courses, including its readings, lecture video subtitles, quizzes, assessments, peer review instructions, and discussion prompts.
    • In addition to AI features, Coursera announced that educators will be able to incorporate virtual reality (VR) components into their courses, “whether that’s practicing a speech in front of a simulated audience or touring the interior of a blood vessel.” Learners will access these immersive experiences via desktop or most VR headsets. [More about new VR capabilities for educators.]
    • Coursera for Campus plagiarism portal. Educators will be able to examine plagiarism metrics and handle learner appeals, ID verification to confirm learners are who they say they are, and enhanced plagiarism detection capabilities on course assessments. “In the coming months, we’ll also be rolling out the ability to improve assessment robustness with hybrid-live proctoring capabilities.”
    • A new grading process, now in Beta, called Quick Grader will enable educators to seamlessly render uploaded assignments to review alongside grading rubrics.
    • Coursera Hiring Solutions, in beta, is a skills-first recruitment platform that matches industry-trained, job-ready talent with companies filling entry-level digital roles.