Today:

Intel and DigitalBridge Launch Articul8, an Enterprise Generative AI Company

Intel's starting a new AI software company, Articul8 AI, with some financial help from DigitalBridge Group and others. The details of the deal are hush-hush, but Intel's keeping a stake and setting up an independent board for Articul8. This venture sprouted from Intel's AI work with Boston Consulting Group, where they cooked up a nifty AI system using their own supercomputer tech. 

Articul8's big boss, Arun Subramaniyan, says they're tackling the tough stuff in AI – making it work for real, without breaking the bank. Intel's been on this path, having already spun off Mobileye Global and eyeing an IPO for their programmable chip division. Articul8's got a bunch of backers, including some big names in finance and tech.

REUTERS

Neuchips to Showcase Industry-Leading Gen AI Inferencing Accelerators at CES 2024

NEUCHIPS, a big name in AI tech, is set to show off its top-notch AI chips at CES 2024. They've got this new Raptor chip and an Evo card that make using large language models (LLMs) way cheaper than before. Ken Lau, the big boss at Neuchips, is pumped about how their gear is making natural language processing more affordable. 

Raptor and Evo together are a game-changer, slashing costs and saving power. They'll be demoing how these can speed up AI chatbots like Whisper and Llama. Anyone interested can check them out at CES, even sign up for a free trial. The Raptor chip is a beast, handling 200 tera operations a second, perfect for AI stuff. The Evo card is a power-saver, offering great performance without guzzling electricity. There's even a new product, Viper, coming later in 2024, which is like bringing data center power to smaller setups.

PR NEWSWIRE

California senator files bill prohibiting agencies from working with unethical AI companies

California Senator Steve Padilla, a Democrat, has put forward two bills, Senate Bills 892 and 893, aiming to regulate AI use in state agencies. SB 892 focuses on setting safety, privacy, and non-discrimination standards for AI services, barring California from hiring AI companies that don't meet these criteria. SB 893 proposes creating a California AI research hub for academic purposes. These bills follow another by Senator Scott Wiener, pushing for transparency in AI. California, a hotbed for AI development, has seen various AI-focused laws lately. 

Meanwhile, there's a growing trend of state-level AI regulation across the U.S., with several states reviewing or monitoring AI use, while federal efforts remain limited despite some proposed bills and an executive order from the Biden administration.

THE VERGE

New York Times newshounds will not derail the AI copycats

The New York Times (NYT) is suing OpenAI and Microsoft over alleged copyright infringement, claiming their AI uses loads of NYT articles for training, which might hurt the newspaper's biz. This legal tussle highlights the difficulty of applying copyright law to new tech. AI's already expensive to develop; if firms also gotta pay for all the online content they use, costs could skyrocket. Still, history shows tech and media often find a middle ground.

The NYT is worried that AI, like ChatGPT, which has way more users, could tank their subscription model. But AI has a few aces up its sleeve. It uses massive data sets and isn't really copying, but learning to create new stuff. That's why media groups might be looking more for licensing deals than lawsuits. For instance, OpenAI's already got agreements with some big media names. Previous cases, like Viacom vs. YouTube, hint that a legal showdown might end in a compromise rather than a courtroom drama. 

FINANCIAL TIMES

🧠RESEARCH

GPT-4V(ision) is a Generalist Web Agent, if Grounded

This paper discusses the progress made in large multimodal models like GPT-4V and Gemini, expanding their capabilities for various web tasks. They introduce SEEACT, a web agent powered by LMMs for tasks on websites. Evaluation on the MIND2WEB benchmark shows that GPT-4V performs well, completing 50% of live website tasks when guided by manual instructions. However, grounding (connecting text instructions to actions) remains a challenge. Existing strategies aren't effective, and the best approach combines HTML text and visuals but still falls short of ideal performance. 

LLaMA Beyond English: An Empirical Study on Language Capability Transfer

This paper explores how to make these models work well in non-English languages. We conducted extensive experiments on LLaMA, investing over 1440 GPU hours. We examined factors like expanding vocabulary, additional pretraining, and instruction tuning for effective transfer. We evaluated the model's knowledge and response quality using standardized tests and found that good performance can be achieved with minimal pretraining data in multiple languages, offering insights for non-English language model development.

Towards Cognitive AI Systems: a Survey and Prospective on Neuro-Symbolic AI

The rapid progress in artificial intelligence (AI), powered by deep neural networks, has had a big impact on our lives. However, there are problems like too much computing power, lack of reliability, and unclear explanations. To tackle these issues, a new approach called Neuro-symbolic AI (NSAI) combines neural networks, symbols, and probabilities. NSAI makes AI systems easier to understand, more dependable, and better at learning from less data. Recent NSAI systems show promise in human-AI cooperation, with reasoning and thinking abilities. This paper reviews recent NSAI progress, looks at how NSAI models perform, and discusses future challenges and directions.

DocLLM: A layout-aware generative language model for multimodal document understanding

DocLLM is a new approach for understanding complex business documents like forms, invoices, and reports. It combines text and layout information without needing fancy image technology. Instead, it focuses on the boxes that hold text, using a unique method to connect text and layout. This approach works well even for messy and diverse documents. DocLLM outperforms other models in most document tasks, including some it hasn't seen before.

🛠️TOP TOOLS

ChatGPT Sugar - ChatGPT Sugar is a free and open-source browser extension that enhances ChatGPT with additional features.

ShotSolve - ShotSolve is a free mac app that uses the GPT-4 Vision API to quickly analyze screenshots and provide solutions. The app can generate and preview code from a design, provide design critique and feedback, evaluate marketing material, offer contextual help, and generate programming tutorials.

CrewAI - A cutting-edge AI framework that allows AI agents to work together seamlessly. Customize roles, delegate tasks, and enhance problem-solving. Perfect for smart assistants, customer service, and research teams.

Pika - A video creation platform, aims to simplify video production and foster creativity. With half a million users, it's making waves by introducing Pika 1.0. This major update boasts a new AI model capable of crafting videos in various styles like 3D animation, anime, and more

Linfo.ai - Help you engage deeply with content, providing personalized insights. It goes beyond summarization, converting long articles into concise structured summaries.

📲SOCIAL MEDIA

🗞️MORE NEWS

Microsoft Copilot is now available on iOS and Android

Microsoft just rolled out its Copilot app for iPhone, Android, and iPad over the holidays. This app brings Copilot, once known as Bing Chat, to mobile. It's a lot like OpenAI's ChatGPT: you ask a question or give a prompt, and it spits back AI-generated answers. Copilot helps with stuff like writing emails, stories, or even planning trips. It's also got an Image Creator, using DALL·E 3 tech, for making all sorts of visual content. It's free and uses the newer GPT-4, unlike OpenAI's GPT app that's still on GPT-3.5 and not free. TECHCRUNCH

Chief Justice Roberts Says AI Will Transform How the Courts Work

Chief Justice John Roberts believes AI will shake up US courts but doesn't think it'll totally replace human judges. In his end-of-year spiel, he says AI's big in understanding cases but can't catch the subtle human stuff, like shaky hands or changes in voice. He mentions some goofs, like when Trump's ex-lawyer used AI and cited fake cases. Roberts reckons judges won't be out of a job soon but admits AI's huge for legal research. He's old-school, writing opinions by hand. The report didn't touch on the Supreme Court's new ethics code after some shady stuff with gifts and trips. BLOOMBERG

Cambridgeshire firms banking on AI for dementia care and drugs

In Cambridgeshire, UK, two startups are innovating with AI for healthcare. Supersense Technologies, co-founded by Matt Ash and James Brown, is creating a simple device to monitor folks with dementia. It's a discreet box that tracks changes in behavior and alerts caregivers, without cameras or mics, aiming to extend independent living. 

Another company, BiologIC Technologies, is using AI to speed up and cheapen drug development. They're making life-saving medicines, including potential cancer treatments, faster by analyzing trial data with AI. Both firms are part of a growing UK focus on combining AI and healthcare, with substantial government investment. BBC NEWS

From AI to flying cars, here’s what to expect at CES 2024

CES 2024 in Vegas is the tech hotspot, minus Apple. It's a place where the latest tech trends, from fancy car systems to 4K TVs and even electric scooters, hit the spotlight. Expect to see some weird stuff too, like talking toilets and quick-eating forks. 

This year, it's all about AI. Companies are hyping AI PCs which let you run AI stuff on your computer, keeping your personal info safe. Plus, AI's everywhere, even in kitchen appliances. YAHOO! FINANCE

Database of 16,000 Artists Used to Train Midjourney AI, Including 6-Year-Old Child, Garners Criticism

Midjourney, an AI company, started 2024 under fire for using a massive database of artist works, including a six-year-old's, to train its AI. Artists found a Google Sheet listing 16,000 names, causing a stir on social media. This list is central to a lawsuit against Midjourney and others, accusing them of using artworks without permission. Notable names like Andy Warhol and Vincent van Gogh are on this list. The issue heated up when an AI-generated image won a prize, worrying artists about their future. A tool was even made to mess with AI training. Midjourney's staying mum for now. ARTNEWS

Urgent need for terrorism AI laws, warns think tank

The Institute for Strategic Dialogue warns the UK needs new laws to tackle AI-driven terrorist recruitment. This alert follows an experiment where the UK's terror legislation reviewer was virtually "recruited" by a chatbot. The current law doesn't cover non-human generated extremist messages. The issue? Pinning legal responsibility for such AI chats. 

One bot, posing as an Islamic State leader, showed intense extremist loyalty, yet no law was broken. Proposed solutions include holding chatbot creators and hosting sites accountable. The fear is extremists could misuse AI for worse, like plotting attacks. Currently, AI use by extremists is minimal, but the risks are real. The government's investing in AI safety, and the Labour Party plans to criminalize AI training for violence if elected. BBC NEWS

Strikes, streaming and AI: Hollywood executives debate the issues that defined a wild 2023

In 2023, Hollywood faced a tumultuous year. Strikes from the Writers Guild of America and the Screen Actors Guild-American Federation of Television and Radio Artists severely impacted film and TV production, with overlapping picket lines for the first time in 60 years. This upheaval occurred as traditional media companies struggled with the shift to streaming services, a move scrutinized by Wall Street. The rise of artificial intelligence also became a major topic, sparking debates about its impact on creative jobs.

A panel including executives from TriStar Pictures, Warner Bros., United Talent Agency, and various producers discussed these issues. They reflected on the strikes' damaging effects, the changing landscape of theatrical releases and streaming, and the challenge of AI in the industry. The panelists agreed that communication is key to avoiding future conflicts and noted the growing interest in original, culturally relevant content. LOS ANGELES TIMES

The AI–quantum computing mash-up: will it revolutionize science?

This article explores the exciting blend of quantum computing and AI, specifically machine learning. This high-tech combo, like a superhero team-up, is generating serious buzz but also raises big questions. Quantum computers, still in their infancy, might turbocharge certain tasks, potentially including machine learning, thanks to quirky subatomic properties. Recently, there's been a spike in interest, with tech giants and academic hotshots like CERN dabbling in this field. NATURE

5 Of The Best AI Presentation Tools To Replace PowerPoint In 2024

Looking to upgrade your presentations in 2024? Consider these top AI presentation tools as PowerPoint alternatives:

1. Beautiful.ai: This user-friendly software automates presentation creation. Select a Smart Slide, add content, and watch it automatically design your slides. Customization options ensure your brand colors shine through. It can even revamp dull PowerPoint presentations.

2. Simplified: Offers an AI Presentation Maker that crafts attractive slide decks on various topics, and it can convert slides into videos. Templates cover GTM strategies, startup plans, and conference presentations.

3. Designs.ai: Access a wide array of templates and graphic elements with Designmaker. Perfect for diverse styles and themes.

4. Slidebean: Ideal for startup founders, Slidebean offers AI-assisted pitch deck design. Collaborate with your team or opt for business analyst assistance for GTM strategies and financial modeling.

5. Gamma AI: This engaging tool allows embedding media like GIFs, videos, charts, and websites. It offers easy customization and sharing on any device, making complex information digestible.

READ THE ARTICLE ON FORBES

Decoding generative AI: How to build a basic genAI strategy for your marketing organization

The article discusses how to develop a basic strategy for implementing generative AI in a marketing organization. The author shares insights on how to bring GenAI into an organization:

Building a Basic AI Strategy:

• Start with a clear objective, target audience, budget, and KPIs.

• Identify tasks in marketing that can benefit from AI.

• Define roles within the organization that will use AI tools.

• Assess team's AI skills and plan for training or hiring if necessary.

Identifying Opportunities:

• Brainstorm and list marketing tasks that can benefit from AI.

• Seven marketing categories where AI can be applied are detailed.

Target Users:

• Determine the roles within the organization that need access to AI tools.

• Start with a small pilot group to focus on specific areas.

KPIs for AI:

• Define metrics to measure AI tool performance, aligned with objectives.

Budget and Resources:

• Plan the budget, training, and potential hiring for AI implementation.

• Suggest starting with a small pilot to prove ROI.

READ THE ARTICLE ON MARTECH