Last Week in AI #212: Bard is underwhelming, ChatGPT gets plug-ins, MidJourney v5 is out, DeepFakes get wrose

Bard takes longer to generate answers than ChatGPT and struggles with easy questions, OpenAI is massively expanding ChatGPT’s capabilities to let it browse the web

Top News

I tested Google Bard. It was surprising -- in a bad way

Summary: The author is underwhelmed by Google's Bard due its limited functions, knowledge, and features. Bard takes longer to generate answers than ChatGPT and Bing Chat, and it struggles to answer straightforward questions. Additionally, unlike Bing Chat, Google Bard does not provide sources for its answers. On the other hand, Google Bard does excel in career-building tools, such as generating resume content and proofreading text, so it remains a useful text editor and generator for tasks like composing cover letters and essays. However, overall, ChatGPT and Bing Chat seem to be more capable chatbots. It should be noted that Google Bard is in its early stages and is expected to improve over time, as it will be updated with newer and more capable models.

Our take: It is surprising that Bard is so much worse than ChatGPT, given that Google was first to create such a chatbot with last year's LaMDA, which is based on very similar techniques to ChatGPT. It may well be that Google is proceeding very carefully to avoid any PR missteps, and as a result Bard is overly constrained. Overall, this is pretty embarassing for Google. Still, as with ChatGPT it is a very powerful tool for all sorts of things, and is likely to improve rapidly.

OpenAI is massively expanding ChatGPT’s capabilities to let it browse the web and more

Summary: OpenAI is introducing plug-in support for ChatGPT, greatly expanding its capabilities and allowing access to live data from the web for the first time. This upgrade will enable ChatGPT to browse the internet and interact with specific websites, potentially transforming it into a comprehensive interface for various services and sites. Initially, 11 plug-ins for external sites such as Expedia, OpenTable, Kayak, Klarna Shopping, and Zapier will be available to a select group of users. These plug-ins allow ChatGPT not only to retrieve real-time information but also to connect to APIs and perform actions on behalf of users. OpenAI has implemented safeguards to address safety and security concerns and will initially prioritize a small number of developers and ChatGPT Plus users for plug-in access.

Our take: This is a big deal. ChatGPT's inability to access new information and interface with various parts of the web had been its biggest limitation, and that is now gone. Moreover, this being a framework for adding arbitrary plug-ins means ChatGPT can be integrated into just about anything, and will likely soon be part of all sorts of websites and applications.

People Aren’t Falling for AI Trump Photos (Yet)

Summary: If you've spent much time on social media recently, you've probably seen a number of photos depicting the arrest of Donald Trump. Images show him running away from police and being detained. These images are not real. With the impending possibility of a Trump indictment, Bellingcat founder Eliot Higgins asked the latest version of Midjourney to illlustrate the spectacle of a Trump arrest. The vivid photos paint a full story arc: from Trump Jr. and Melania Trump screaming at officers to Trump's eventual escape from prison. Only the scene of Trump's arrest went viral—but despite widespread and not-entirely-new concerns about misinformation, most people who saw the images were not fooled. But we're not all immune to misinformation. Chris Moran, head of editorial innovation at The Guardian was contacted by a researcher asking why the paper had deleted an article from its archive—this article did not exist, but was hallucinated entirely by ChatGPT. As The Atlantic concludes, we can't prognosticate much except to say that things are likely to get weird.

Our Take: I recently spoke to Irene Solaiman (episode isn't released yet), who directs AI policy at HuggingFace, about concerns like deepfake-powered misinformation campaigns. While some of the mass misinformation campaigns researchers worried about have not come to pass, red-teaming and understanding what negative scenarios for modern technologies look like is a useful exercise. There's no doubt that AI systems could complicate our sense of reality, but it's hard to know exactly what that will look like. Exercises like understanding the implications of and possible mitigation strategies for misinformation campaigns and similar problems is a useful exercise. If these things do come to pass, even if not in the form we expected, someone will have thought about how they will play out and what to do.

Other News

Research

Learning to grow machine-learning models - " New LiGO technique accelerates training of large machine-learning models, reducing the monetary and environmental cost of developing AI applications."

Researchers develop a four-wheeled, two orthogonal axes mechanism robot to maintain plants grown under solar panels - "Synecoculture is a new agricultural method advocated by Dr. Masatoshi Funabashi, senior researcher at Sony Computer Science Laboratories, Inc."

Mind-control robots a reality - "Researchers from the University of Technology Sydney (UTS) have developed biosensor technology that will allow you to operate devices, such as robots and machines, solely through thought control."

New virtual testing environment breaks the 'curse of rarity' for autonomous vehicle emergency decision-making - "The push toward truly autonomous vehicles has been hindered by the cost and time associated with safety testing, but a new system developed at the University of Michigan shows that artificial intelligence can reduce the testing miles required by 99.99%."

MIT Researchers Introduce LiGO : A New Technique that Accelerates Training of Large Machine-Learning Models, Reducing the Monetary and Environmental Cost of Developing AI Applications - "The Computer Science and Artificial Intelligence Laboratory (CSAIL) suggests using pre-trained, smaller language models to boost the effectiveness of these training approaches at a reduced cost and time commitment."

Researchers from UC Berkeley and Deepmind Propose SuccessVQA: A Reformulation of Success... - "In order to achieve the best possible performance accuracy, it is crucial to understand whether an agent is on the right or preferred track during training."

Scientists are using machine learning to forecast bird migration and identify birds in flight by their calls - "Last year, researchers at the Cornell Lab of Ornithology published an automated system that uses machine learning to detect and identify nocturnal flight calls."

Applications

Microsoft brings OpenAI’s DALL-E image creator to the new Bing - "Microsoft today announced that its new AI-enabled Bing will now allow users to generate images with Bing Chat. This new feature is powered by DALL-E, OpenAI’s generative image generator."

Tavus taps generative AI to power personalized videos with voice and face cloning - "Generative AI is already looking like the major tech trend of 2023."

Microsoft subsidiary Nuance is using GPT-4 for a new physician notes app - "Microsoft-owned AI company Nuance is leveraging GPT-4 to power new software designed to ease the burden of clinical documentation for physicians. Nuance, acquired last year by Microsoft for $19."

Midjourney V5 is Out Now – Next Steps in Photorealistic Experience with AI Art - "A recent breakthrough in AI, you might have missed: the highly awaited Midjourney V5 is out now. The independent research lab has just released their latest version of the famous AI art generator."

If You Can Say It, Now You Can See It: RunWay’s Latest Artificial Intelligence Tool Can Generate Videos With Nothing But Words - "Runway, an artificial intelligence (AI) platform, has recently released its latest software, Gen-2, which can create full videos from text descriptions. Gen-2 was developed after Runway’s previous model, Gen-1, which used existing video data to make new videos."

Business

Agility’s Latest Digit Robot Prepares for its First Job - "This morning at the ProMat conference in Chicago, Agility Robotics is introducing the latest iteration of Digit, their bipedal multipurpose robot designed for near-term commercial success in warehouse and logistics operations."

Microsoft Business Chat is like the Bing AI bot but as a personal assistant - "One of the new Copilot AI features coming to Microsoft 365 apps and services is dubbed Business Chat."

Nimble Robotics raises $65M to scale up its autonomous logistics fulfillment network - "Nimble Robotics Inc., an autonomous logistics startup that’s aiming to reinvent the fulfillment center for small and medium-sized businesses, said today it has closed on a $65 million round of funding."

Numbers Station raises $17.5M to bring AI to your data stack - "When you think of foundation models and what they can accomplish, chances are you are thinking of text and image creation. Numbers Station, however, is taking these models in a very different direction. The startup, which is announcing a $17."

Former head of Google China joins ChatGPT frenzy by starting own venture - "Lee Kai-fu, a prominent Chinese venture capitalist and former president of Google China, is the latest high-profile tech executive to jump on the ChatGPT bandwagon, with plans to establish an artificial intelligence (AI) company that goes beyond “a Chinese version of ChatGPT”."

OpenAI co-founder on company’s past approach to openly sharing research: ‘We were wrong’ - "Yesterday, OpenAI announced GPT-4, its long-awaited next-generation AI language model."

NVIDIA's big AI moment is here - "When NVIDIA's founder and CEO Jensen Huang waxed poetic about artificial intelligence in the past, it mostly felt like marketing bluster — the sort of lofty rhetoric we've come to expect from an executive with a never-ending supply of leather jackets."

The top 10 AI mobile apps have already pulled in over $14 million this year - "Consumer demand for AI chatbot experiences has been funneling millions of dollars into mobile apps advertising their association with ChatGPT or OpenAI technologies. According to a new analysis of the AI app ecosystem from analytics provider data."

Chatbot Start-Up Character.AI Valued at $1 Billion in New Funding Round - "Character.AI, a 16-month-old start-up that builds online chatbots, said on Thursday that it had raised $150 million in a recent funding round that valued the company at $1 billion. Character."

Concerns

AI can draw hands now. That’s bad news for deep-fakes. - "AI image-makers have been notoriously bad at rendering hands. Now, they’re getting better and could make deep fakes harder to spot."

Samsung’s Moon Shots Force Us to Ask How Much AI Is Too Much - "Have you heard the moon conspiracy theory? No, not the one about the moon landings. It’s about the Samsung Galaxy S23 Ultra, and the theory it fabricates pictures of the moon, creating images that are far more detailed than the camera itself can actually capture."

Voice system used to verify identity by Centrelink can be fooled by AI - "A voice identification system used by the Australian government for millions of people has a serious security flaw, a Guardian Australia investigation has found."

In San Francisco, some people wonder when AI will kill us all - "Audrey Kim is pretty sure a powerful robot isn't going to harvest resources from her body to fulfill its goals. But she's taking the possibility seriously."

OpenAI’s policies hinder reproducible research on language models - "Researchers rely on ML models created by companies to conduct research. One such model, OpenAI's Codex, has been used in about a hundred academic papers On Monday, OpenAI announced that it would discontinue support for Codex by Thursday."

Robot recruiters: can bias be banished from AI hiring? - "Michael Scott, the protagonist from the US version of The Office, is using an AI recruiter to hire a receptionist. Guardian Australia applies."

AI expert Meredith Broussard: ‘Racism, sexism and ableism are systemic problems’ - "Meredith Broussard is a data journalist and academic whose research focuses on bias in artificial intelligence (AI). She has been in the vanguard of raising awareness and sounding the alarm about unchecked AI."

Analysis

GPT-4 and professional benchmarks: the wrong answer to the wrong question - "OpenAI may have tested on the training data. Besides, human benchmarks are meaningless for bots."

A Vision for the Future: How Computer Vision is Transforming Robotics - "The goal of computer vision research is to teach computers to recognize objects and scenes in their surroundings. As robots need to be able to pick up on their surroundings and adapt accordingly, this is a crucial skill for the field."

We Need to Talk (Just as Soon as I Consult ChatGPT) - "Faced with challenging situations in parenting, romance or work, some are getting by with a little help from their A.I. friends. Todd Mitchem kept struggling to have honest, productive conversations with his son. “He’s 15,” Mr. Mitchem, 52, said with a laugh."

A.I. Can’t Write My Cat Story Because It Hasn’t Felt What I Feel - "For a few years, I’ve been trying to write a story about a cat. A.I."

OpenAI Research Says 80% of U.S. Workers' Jobs Will Be Impacted by GPT - "As large language models like OpenAI's GPT-4 become more advanced and able to write, code, and do math with more accuracy and consistency, it won’t be long before AI makes its way into the workplace as a daily-use tool."

Expert Opinions

Bill Gates: AI is most important tech advance in decades - "Microsoft co-founder Bill Gates says the development of artificial intelligence (AI) is the most important technological advance in decades. In a blog post on Tuesday, he called it as fundamental as the creation of the microprocessor, the personal computer, the Internet, and the mobile phone."

Explainers

Prompt Engineering - "Prompt Engineering, also known as In-Context Prompting, refers to methods for how to communicate with LLM to steer its behavior for desired outcomes without updating the model weights."

Emergent Abilities of Large Language Models - "Emergence can be defined as the sudden appearance of novel behavior. Large Language Models apparently display emergence by suddenly gaining new abilities as they grow. Why does this happen, and what does this mean?"

---

Copyright © 2023 Skynet Today, All rights reserved.