Last Week in AI #236: Pentagon's AI fleet against China, Imbue's $200M boost for reasoning AI, AI sex worker ads invade social media, Anthropic's paid plan for Claude, and more!

Pentagon plans vast AI fleet to counter China, Imbue raises $200M to build AI models that can ‘robustly reason’, Ads for AI sex workers are flooding Instagram and TikTok

Top News

Pentagon plans vast AI fleet to counter China, Wall Street Journal reports

The Pentagon is planning to invest heavily in a large-scale network of AI-powered technology, drones, and autonomous systems over the next two years to counter increasing threats from China and other adversaries. Kathleen Hicks, the deputy secretary of defense, unveiled plans to allocate hundreds of millions of dollars towards the development of thousands of AI systems operating in air, land, and sea, emphasizing their need to be "small, smart, cheap." The move is in response to China's swiftly advancing military and the Pentagon's own bureaucratic delays in deploying state-of-the-art systems. The strategy might take inspiration from Task Force 59, the U.S. Navy's drone and sensor network aimed at monitoring Iran. While the defense department is keen on integrating technologies like sun-powered autonomous systems, space-based systems, and anti-missile systems, questions about funding and integration with existing military infrastructure persist. The Pentagon intends to finalize an initial investment list by year-end.

Imbue raises $200M to build AI models that can ‘robustly reason’

AI research lab Imbue, formerly known as Generally Intelligent, has secured $200 million in a Series B funding round, bringing its total funding to $220 million and valuing the company at over $1 billion. The funding will be used to accelerate the development of AI systems that can reason and code, with the ultimate goal of creating practical AI agents that can work safely in the real world. Imbue's approach involves designing AI models that can solve tasks in complex 3D worlds, with a recent shift towards developing models that can code. The company believes that robust reasoning is crucial for effective AI agents and that code can improve reasoning and is an effective way for models to take actions on a machine. Imbue is also investing in its own AI and machine learning tooling and conducting research into understanding the learning process in large language models.

Ads for AI sex workers are flooding Instagram and TikTok

Sexually explicit advertisements for AI chatbots have been proliferating on social media platforms like Instagram, Facebook, and TikTok, despite these platforms' strict policies against sexualized content. The ads, run by dozens of tech startups, promote not-safe-for-work experiences with AI-generated characters, often using popular children's TV characters or anime-style images. NBC News found 35 app developers running such ads on Meta-owned platforms and 14 on TikTok. The rise of these ads is part of a larger trend to capitalize on the growing interest in AI, following the success of OpenAI's ChatGPT product. Critics argue that these ads reflect a double standard, as real human sex workers are not allowed to monetize their images on these platforms. Both Meta and TikTok have begun removing these ads after being contacted by NBC News, but questions remain about how they bypassed the platforms' filters in the first place.

Anthropic’s Claude AI chatbot gets a paid plan for heavy users

Anthropic, an AI firm supported by Google, has introduced a paid version of its Claude chatbot for heavy users in the US and UK. The Claude Pro plan, priced at $20 or £18 per month, provides priority access during peak times, early access to new features, and the ability to send more messages. The key advantage of this plan is that it allows five times more usage than the free tier, enabling users to send more messages in less time. The company promises a minimum of 100 messages every eight hours based on Claude's capacity, with a warning system in place when only 10 messages are left, and a reset every eight hours.

Other News

Applications

Artificial intelligence allows paralysis patient to speak for first time in 18 years - Researchers at the University of California have developed a device that uses AI and surgically implanted electrodes to restore communication in paralyzed patients unable to speak, offering a breakthrough in the fields of artificial intelligence, neural engineering, and medicine.

Call of Duty partners with Modulate to use AI to fight toxicity in voice chat - Call of Duty has partnered with Modulate to use AI-based voice chat moderation technology to combat toxic behavior and enforce the Call of Duty Code of Conduct.

How generative AI helped train Amazon One to recognize your palm - Generative AI was used to train Amazon One, a contactless payment system that uses palm recognition, by creating millions of synthetic images of palms to boost accuracy and speed up the training process.

The latest canvas for Refik Anadol’s AI-generated art? The new Sphere in Las Vegas - Digital artist Refik Anadol will use the world's largest LED screen, located on the exterior of the Sphere in Las Vegas, as a canvas for his AI-generated art piece "Machine Hallucinations: Sphere," which draws on outer space, nature, and urban environments.

Slack AI' will summarize your work chat starting this winter - Salesforce is gearing up to pilot Slack AI, a new feature that can generate channel highlights, summaries for threads, and elaborate answers to search queries on the platform.

Chinese AI chatbots want to be your emotional support - Chinese AI chatbots, such as Baidu's Ernie Bot, are gaining popularity among users due to their interactive features and user-generated content, which helps to keep users engaged and interested.

Roblox’s new AI chatbot will help you build virtual worlds - Roblox has announced a new conversational AI assistant called Roblox Assistant that can help creators build virtual environments and experiences for the popular social app, with plans to expand its capabilities to include generating gameplay and 3D models in the future.

Revolution Software is using their own AI technology to remake Broken Sword - Revolution Software is using their own AI technology to upscale the animations in the remake of Broken Sword - The Shadow of the Templars, allowing the studio to focus on adding expressions and character to the game.

The Halo Effect: AI Deep Dives Into Coral Reef Conservation - Researchers from the University of Hawaii have developed an AI-based surveying tool that uses deep learning models and high-resolution satellite imagery to monitor the health of coral reefs by spotting and tracking coral reef halos, which could revolutionize real-time coral reef monitoring and global conservation efforts.

SoTaNa: The Open-Source Software Development Assistant - SoTaNa is an open-source software development assistant that utilizes ChatGPT and a parameter-efficient fine-tuning approach to enhance the open-source foundation model, LLaMA, in order to answer Stack Overflow questions and perform code summarization and generation.

Ghostwriter Returns With an A.I. Travis Scott Song, and Industry Allies - Ghostwriter, the anonymous musical creator, released an unauthorized track that used AI voice effects to mimic popular artists, prompting discussions among industry figures about the implications of AI in creativity and the music business.

Business

ChatGPT creator OpenAI is reportedly earning $80M a month - OpenAI, the creator of ChatGPT, is reportedly earning $80 million a month and is on track to generate over $1 billion in annual revenue, defying previous expectations.

Intuit launches generative AI–powered digital assistant for small businesses and consumers - Intuit has launched Intuit Assist, a generative AI-powered digital assistant that provides personalized recommendations and assistance to small businesses and consumers across its platform and products, including TurboTax, Credit Karma, QuickBooks, and Mailchimp.

Qualcomm focuses on AI and auto as Nvidia takes over as world’s biggest fabless chip company - Qualcomm is positioning itself as a key player in the AI space, with new in-vehicle generative AI capabilities, expansion into two-wheelers, and a partnership with Amazon Web Services, as it aims to run generative AI models locally and improve performance and cost.

AI startup Speak raises $16 million in latest funding led by Lachy Groom - AI language learning platform Speak raises $16 million in funding led by Lachy Groom, bringing the total amount raised to $63 million and enabling expansion into international markets.

AI chip startup d-Matrix raises $110 mln with backing from Microsoft - Silicon Valley-based AI chip startup d-Matrix raises $110 million in funding, with backing from Microsoft, to develop chips optimized for generative AI applications and compete with Nvidia in the AI chip market.

Zoom rebrands existing — and intros new — generative AI features - Zoom is rebranding and updating its AI-powered features, including the generative AI assistant formerly known as Zoom IQ, in order to stay competitive in the videoconferencing market.

Gizmodo’s owner shuts down Spanish language site in favor of AI translations - Gizmodo's owner, G/O Media, has shut down its Spanish-language site and is now using AI to automatically translate articles, despite previous promises that AI would only supplement human writing.

Zoom Is Jumping on the AI Chatbot Bandwagon - Zoom plans to introduce an AI chatbot called AI Companion to its videoconferences and chats, which will allow users to ask for help with office tasks and interact with the chatbot in real-time, but this move may reignite concerns over data privacy and the type of data used to train the AI.

China's Ant Group unveils finance AI model as race heats up - Ant Group, China's biggest fintech firm, has unveiled a finance-specific AI model and started testing consumer and professional apps for the product, joining other tech giants in the race to deploy AI in heavily regulated China.

Watch out, Midjourney! Ideogram launches AI image generator with impressive typography - Ideogram, a new generative AI image startup founded by former Google Brain researchers, has launched with $16.5 million in seed funding and aims to solve the problem of reliable text generation within AI-generated images.

AI reading coach startup Ello raises $15M to bolster child literacy - AI reading coach startup Ello has raised $15 million in funding to expand its product and improve child literacy by leveraging artificial intelligence and child speech recognition technology.

Morgan Stanley to launch AI chatbot to woo wealthy - Morgan Stanley is set to launch an AI chatbot developed with OpenAI to assist financial advisers by quickly finding research or forms and potentially creating meeting summaries and follow-up emails.

ThetaRay nabs $57M for AI tools to ID and fight money laundering - ThetaRay, an Israeli startup, has raised $57 million in funding for its AI-based anti-money laundering platform that automatically scans and identifies illicit transaction activity at banks and other financial institutions, with customers including Santander Bank and Travelex.

Sapeon raises $46m for AI chips - Sapeon, a spin out of SK Telecom, sk Hynix, and SK Square, has raised $46m for its AI chip and software development, with plans to provide AI services ranging from hardware to software and AI-based services.

Nvidia enters AI partnership with India's Tata Group - Nvidia and Tata Group have formed a partnership to provide computing infrastructure and platforms for AI development.

ChatGPT’s move to lure businesses into the A.I. waters gets mixed reactions - OpenAI's release of the enterprise tier for ChatGPT has received mixed reactions from businesses, with some finding the improved security features and peace of mind necessary to integrate the tool, while others have concerns about privacy, security, and compliance, and are holding out on adopting the technology.

Cobot arm maker Franka Emika files for insolvency - German cobot arm maker Franka Emika has filed for insolvency due to shareholder differences, with the preliminary insolvency administrator working to secure wages for employees and find a solution for the company's future.

QuantHealth brings its AI-informed clinical drug trials to the US with $15M round - QuantHealth, a Tel Aviv-based startup, has raised $15 million in funding to support its AI-powered platform for drug discovery, which aims to reduce risks, optimize clinical trials, and predict outcomes for patients in a clinical trial.

Bybit debuts AI-powered 'TradeGPT' for market analysis and data driven Q&A - Bybit has released an AI-powered trading assistant called TradeGPT that uses language-based technology to provide real-time market analysis and answer user questions, offering market strategies and products for discussion.

No VC money, no problem for this AI startup - An AI startup proves that it doesn't need venture capital funding to succeed.

Baidu CEO says more than 70 large AI language models released in China - Baidu CEO announces the release of over 70 large AI language models in China, including an improved version of their AI chatbot with faster processing speed and improved efficiency.

Tencent releases AI model for businesses as competition in China heats up - Tencent is launching its AI model "Hunyuan" for business use, integrating its capabilities with existing products for video conferencing and social media, as competition in China heats up with Baidu's recent AI-powered applications.

Powered by A.I., Company Aims to Make Selling Easier for Retailers - An AI-powered platform called Lily AI is helping fashion retailers improve their online shopping experience by accurately matching merchandise to colloquial terms that shoppers type in the search bar.

Nasdaq gets SEC nod for first exchange AI-driven order type - Nasdaq has received approval from the SEC to launch an AI-driven order type that aims to increase the efficiency of the stock market by speeding up order matching and minimizing market impact.

TSMC warns AI chip crunch will last another 18 months - TSMC warns that the shortage of AI chips, including Nvidia's top specced GPUs, will continue for another 18 months due to a lack of advanced packaging capacity.

Concerns

Unraveling the Ethical Complexities of A.I. - Artificial intelligence has rapidly evolved and raises serious ethical questions about aligning with human values, potential job displacement, bias and fairness, privacy concerns, AI's role in warfare and surveillance, and the need for responsible development and regulation.

Are AI models doomed to always hallucinate? - Large language models (LLMs) like OpenAI’s ChatGPT have a tendency to invent "facts" and make mistakes, a phenomenon known as hallucination, due to the way they are developed and trained, but there are ways to reduce hallucinations depending on how the LLM is trained and deployed, such as curating a high-quality knowledge base or using reinforcement learning from human feedback.

China suspected of using AI on social media to sway US voters, Microsoft says - Microsoft researchers have discovered a network of fake, Chinese-controlled social media accounts that are suspected of using artificial intelligence to influence US voters.

Why You Should be Skeptical of AGI - The article discusses the skepticism surrounding claims of achieving artificial general intelligence (AGI) and highlights the lack of consensus on its definition and arrival, as well as the limitations of current AI systems in terms of reasoning and human-like intelligence.

A.I.’s un-learning problem: Researchers say it’s virtually impossible to make an A.I. model ‘forget’ the things it learns from private user data - It is virtually impossible to remove private user data from an AI model without resetting the model entirely, posing a significant challenge in terms of privacy and data protection.

Trusting A.I.-written mushroom hunting guides sold on Amazon could get you killed. But like deadly fungi, identifying them is tricky - Mushroom hunting guides written by AI are being sold on Amazon and other retailers, posing a risk to readers who may receive deadly advice.

Behind the AI boom, an army of overseas workers in ‘digital sweatshops’ - Thousands of workers in the Philippines are laboring in "digital sweatshops," annotating data for AI models for San Francisco start-up Scale AI, often without receiving proper payment.

AI Drake song ‘not eligible’ for a Grammy, Academy CEO says - The AI Drake song "Heart on My Sleeve" will not be eligible for a Grammy because the vocals were not legally obtained, cleared by the label or artists, and the song is not commercially available, according to Recording Academy CEO Harvey Mason Jr.

Research

New Open-Source ‘Falcon’ AI Language Model Overtakes Meta and Google - Falcon 180B, an open-source large language model with 180 billion parameters, has surpassed previous models in scale and benchmark performance, rivaling commercial models like Google's PaLM-2, making it an exciting development for open-source AI.

UAE's G42 launches open source Arabic language AI model - A group of engineers and researchers, in collaboration with Cerebras Systems, has released an advanced Arabic language software called Jais, which is an open source model that can power generative AI applications and contains 13 billion parameters.

Meta releases a data set to probe computer vision models for biases - Meta has released a new AI benchmark called FACET, which is designed to evaluate the "fairness" of AI models that classify and detect things in photos and videos, including people, and aims to enable researchers and practitioners to perform benchmarking to better understand the disparities present in their own models and monitor the impact of mitigations put in place to address fairness concerns.

Scaling Autoregressive Multi-Modal Models: Pretraining and Instruction Tuning - The article discusses the scaling of autoregressive multi-modal models through pretraining and instruction tuning.

Sequential Dexterity: Chaining Dexterous Policies for Long-Horizon Manipulation - A system based on reinforcement learning is presented that chains multiple dexterous policies to achieve long-horizon task goals, demonstrating generalization capability to novel object shapes and zero-shot transfer to a real-world robot.

Scientists used machine learning to perform quantum error correction - Scientists from the RIKEN Center for Quantum Computing used machine learning to perform error correction for quantum computers, finding methods with low device overhead and high error-correcting performance, bringing them a step closer to the successful implementation of quantum error correction in experiments.

Researchers from Microsoft Introduce Hydra-RLHF: A Memory-Efficient Solution for Reinforcement Learning with Human Feedback - Researchers from Microsoft have introduced Hydra-RLHF, a memory-efficient solution for reinforcement learning with human feedback, which reduces memory usage and improves training batch size, resulting in faster per-sample latency for PPO.

LLM Benchmarks: What Do They All Mean? - The article provides a curated and categorized list of the most common benchmarks used to measure the performance of large language models (LLMs) in various tasks, including natural language processing, general knowledge and common sense, problem-solving, and coding.

SyncDreamer: Generating Multiview-consistent Images from a Single-view Image - The article discusses the generation of multiview-consistent images from a single-view image using SyncDreamer.

Pre-gated MoE: An Algorithm-System Co-Design for Fast and Scalable Mixture-of-Expert Inference - A proposed Pre-gated MoE system addresses the computational and memory challenges of large language models (LLMs) by using a novel pre-gating function, reducing GPU memory consumption while maintaining high performance.

Perception, performance, and detectability of conversational artificial intelligence across 32 university courses - Large language models like ChatGPT can perform at a comparable or even superior level to students in university courses, particularly in disciplines like Data Structures, Public Policy, and Synthetic Biology, but there are variations in performance based on the level of knowledge and cognitive process required; educators and students have mixed opinions on the ethics and impact of using ChatGPT in education, with some believing it should be acknowledged and others considering it unethical or worrying about job displacement; there is a consensus that ChatGPT will increase

Scientists teach a neural network to recognize PC users' fatigue - Scientists have created a database of eye movement strategies to train neural network models that can accurately detect fatigue in PC users, with the goal of developing high-precision systems for tracking functional states to ensure safety in various industries.

The Belebele Benchmark: a Parallel Reading Comprehension Dataset in 122 Language Variants - Belebele is a multiple-choice machine reading comprehension dataset that spans 122 language variants, allowing for the evaluation of text models in high-, medium-, and low-resource languages and providing insights into the multilingual capabilities of NLP systems.

RLAIF: Scaling Reinforcement Learning from Human Feedback with AI Feedback - RLAIF is a method that scales reinforcement learning by combining human feedback with AI feedback.

Scaling Autoregressive Multi-Modal Models: Pretraining and Instruction Tuning - The article discusses the scaling of autoregressive multi-modal models through pretraining and instruction tuning.

Analysis

How We Chose the TIME100 Most Influential People in AI - TIME has chosen the TIME100 Most Influential People in AI, highlighting industry leaders, innovators, and individuals grappling with ethical questions surrounding AI.

Datasets as Imagination - The article discusses the potential for artists to lead the development of datasets for AI art, highlighting the exploitative nature of current datasets and proposing a future where artist collectives curate datasets with intention and contextual specificity, providing new opportunities for creativity and income.

AI researchers' challenges: atomic analogies and strained institutions - The pressures on the AI community to compare themselves to the Manhattan Project and the role of social media algorithms in research distribution are contributing to a confusing and strained moment in time for AI and those working with large language models.

AI is not creative. Yet. - AI is not inherently creative because it lacks the ability to break rules and make conscious decisions, and its output is dependent on human prompts and intentions.

How will artificial intelligence change the value of human skillsets?   - Artificial intelligence will impact the value of human skillsets, but the article does not provide any information on how it will do so.

Policy

Inside Elon Musk's Struggle for the Future of AI - Elon Musk's concern about the dangers of artificial intelligence led him to invest in DeepMind, co-found OpenAI, and start his own AI company called xAI, with the goal of making AI safe and preserving human consciousness.

2 Senators Propose Bipartisan Framework for A.I. Laws - Senators Richard Blumenthal and Josh Hawley are proposing a comprehensive framework to regulate artificial intelligence, including requirements for licensing and auditing, the creation of an independent federal office, liability for privacy and civil rights violations, and data transparency and safety standards.

UK government sets out AI Safety Summit ambitions - The UK government has outlined its objectives for the AI Safety Summit, which aims to address the risks and potential benefits of frontier AI and foster international collaboration on AI safety.

Britain sets priorities for November global AI safety summit - Britain has outlined its objectives for a global AI safety summit, which will focus on understanding the risks posed by AI and how national and international frameworks can support regulation of the technology.

The UK releases key ambitions for global AI summit - The UK has released its five "ambitions" for the upcoming global AI safety summit, which will focus on risks and policy surrounding AI technology.

Congress to hold new AI hearings as it works to craft safeguards - Congress is holding three hearings on artificial intelligence, including discussions on legislation to mitigate the dangers of AI, establishing guardrails, protecting against AI harms, and ensuring transparency and accountability.

Microsoft says it will take the heat if Copilot AI users get sued - Microsoft will assume legal responsibility and defend customers if they get sued for copyright infringement while using its AI Copilot services, as the company aims to address concerns from copyright holders and ensure that content needed to train AI models is not monopolized.

How one elite university is approaching ChatGPT this school year - Yale University is embracing the use of ChatGPT in the classroom, encouraging educators to reconsider what they are asking students to learn and how it can be justified in the face of AI capabilities.

U.S. Copyright Office Invites Public To Comment On AI - The U.S. Copyright Office has launched a study on artificial intelligence (AI) and copyright law, inviting the public to comment on various policy issues related to AI and intellectual property, including questions about AI training, transparency, copyright liability, and more.

Artists sign open letter saying generative AI is good, actually - Artists argue in an open letter to Congress that generative AI is a valuable tool for creativity and should be regulated with input from the creative community, despite concerns about IP theft and exploitation.

Prosecutors in all 50 states urge Congress to strengthen tools to fight AI child sexual abuse images - Prosecutors in all 50 states are urging Congress to study how artificial intelligence can be used to exploit children through pornography and come up with legislation to further guard against it.

US court rules that artificial intelligence generated artwork cannot be copyrighted - US court rules that artificial intelligence generated artwork cannot be copyrighted, establishing legal limitations on protections for AI generated works and raising questions for copyright laws as artists incorporate AI into their artworks.

Fun

AI Took the Stage at the World's Largest Arts Festival. Here's What Happened - AI took center stage at the Edinburgh Fringe festival, with performances exploring the nuances of AI through theater, stand-up comedy, dance, and clowning, highlighting the ongoing debate about the impact of AI on creativity and artistry.

---

Copyright © 2023 Skynet Today, All rights reserved.