Sponsored by

Today:

Visual-based newsletter on business and tech

We explain the latest business, finance, and tech news with visuals and data. 📊

All in one free newsletter that takes < 5 minutes to read. 🗞

Save time and become more informed today.👇

Google Gemini Advanced | Gemini Ultra 1.0 Release Date and Bard Shutting Down

Google's gearing up to shake things up with a major update next week, ditching the Google Bard name in favor of "Gemini". The big scoop? They're launching "Gemini Advanced", basically a supercharged Bard with their latest AI tech, Ultra 1.0. 

The buzz is all about pushing the envelope in AI, with Gemini aiming to outdo competitors, including a special nod to the Galaxy S24 and its AI smarts. 

WATCH THE VIDEO ON YOUTUBE

Google’s CEO faces employee questions about layoffs

In a recent meeting at Google, CEO Sundar Pichai addressed concerns about layoffs and the company's AI strategy. The discussion also touched on ByteDance's CEO pushing for "urgency" and Meta's significant developments this week.

The details were shared by Alex Heath, a tech industry veteran. Pichai's comments on these matters were obtained through audio recordings, providing insight into Google's current challenges and strategies.

THE VERGE

MIT and IBM Find Clever AI Ways Around Brute-Force Math 

Researchers have devised a novel approach to solving complex equations that underpin various fields of science and engineering. These equations, like those modeling the flow of air over aircraft wings or the behavior of pollutants in the atmosphere, traditionally demanded high-precision numerical methods, which are time-consuming and resource-intensive. Current alternatives, such as neural networks trained on numerical solver data, also require extensive training data.

The new method, called physics-enhanced deep surrogate (PEDS) models, combines physics simulators with neural networks to achieve accuracy three times greater than other neural networks, requiring just 1,000 training points, significantly reducing data demands for solving these equations.

IEEE SPECTRUM

Hugging Face makes it easier to create its custom chatbots.

Hugging Face just leveled up its game. Now, you can whip up your own chatbots in just two clicks with their Hugging Chat Assistant. It's as simple as pie. And guess what? Your chatbots can go public, too.

Hugging Face's tech lead, Philipp Schmid, even threw some shade at OpenAI's GPTs, saying it's like a walk in the park compared to them. Plus, you can use any fancy-sounding LLM like Llama2 or Mixtral.

THE VERGE

China rushes to homegrown AI chips as Nvidia cutoff expands

Chinese artificial intelligence companies are turning to domestic suppliers like Huawei Technologies as U.S. export restrictions make it increasingly difficult to access key Nvidia chips. Previously, Nvidia held a dominant position in China's AI semiconductor market, but export restrictions have limited its availability in the country.

Chinese tech giants like Alibaba, Tencent, and Baidu are now seeking alternative sources for AI chips. Huawei, despite facing U.S. sanctions, is marketing its Ascend AI chipset as a potential alternative, although its performance is currently considered inferior to Nvidia chips. Other Chinese chipmakers are also entering the market, but mass production remains a challenge.

NIKKEI

🧠RESEARCH

Hugging Face Assistants

Hugging Face has launched Hugging Chat assistants, enabling users to create custom chatbots similar to OpenAI's GPT models. These assistants offer a range of applications, from productivity tools to quirky companions, showcasing the versatility of AI in engaging and innovative ways. You can design your unique chat assistant tailored to your needs or interests. 

AnimateLCM: Accelerating the Animation of Personalized Diffusion Models and Adapters with Decoupled Consistency Learning

"AnimateLCM" cranks up video creation with AI, making high-quality vids faster by splitting image and motion learning. It plugs into existing tools for more control, showing top-notch results in tests. They're sharing how it's done, promising to release the code and tech details for everyone to use. 

AToM: Amortized Text-to-Mesh using 2D Diffusion

AToM introduces a speedy way to turn text into 3D meshes, making it faster and cheaper than before. Unlike older methods that took a while and gave you something less useful, AToM pops out detailed, textured 3D shapes in under a second and works on stuff it hasn't even seen before. It's a big leap over the competition, delivering clearer and better-quality 3D creations. 

OLMo: Accelerating the Science of Language Models

OLMo aims to supercharge language model science by offering full transparency. Unlike others that only share parts, OLMo gives away everything: model weights, training data, and even the code for training and testing. This move is set to fuel open research, inviting a fresh wave of advancements in understanding and improving language models. It addresses the need for open access to powerful language tools for scientific study, including examining biases and potential risks.

Dolma: an Open Corpus of Three Trillion Tokens for Language Model Pretraining Research

"Dolma" unveils a massive, open-source English corpus of three trillion tokens to boost language model pretraining research. It's a diverse blend of web content, scientific papers, code, books, social media, and more. By sharing their data curation toolkit, the team encourages experimentation and replication of their work. This corpus, which trained OLMo, sheds light on data curation's impact on model capabilities and limitations. 

SpatialVLM: Endowing Vision-Language Models with Spatial Reasoning Capabilities

"SpatialVLM: Endowing Vision-Language Models with Spatial Reasoning Capabilities" addresses the challenge of spatial reasoning in Visual Question Answering (VQA) and robotics. It aims to enhance the spatial reasoning capability of Vision Language Models (VLMs) by training them with internet-scale spatial reasoning data. The paper presents a system that develops an automatic 3D spatial VQA data generation framework capable of generating 2 billion VQA examples on 10 million real-world images. Various factors in the training process, including data quality and model architecture, are investigated. The resulting VLM shows improved spatial reasoning abilities, enabling novel applications in spatial reasoning and robotics.

🛠️TOP TOOLS

Aspen - free API-testing tool designed specifically for macOS, offering a native app experience without the need for user login.

PDFtoChat - Chat with your PDFs in seconds

Talently - AI-based interviewing platform that automates live, conversational interviews and real-time evaluations, streamlining the recruitment process by providing objective candidate scoring and role-specific interviewing, including live coding assessments.

TextReader - free text-to-speech tool with realistic AI voices, transforming written text into lifelike speech quickly. 

Grantable -  AI-powered grant writing assistant that simplifies and streamlines the grant writing process, helping users craft proposals efficiently and effectively. 

Fireflies AI - AI-based tool that records, transcribes, and summarizes meetings. It integrates with various meeting platforms, offers AI-powered search, facilitates collaboration, and provides analysis on conversation intelligence.

📲SOCIAL MEDIA

🗞️MORE NEWS

UK government urged to adopt more positive outlook for LLMs to avoid missing ‘AI goldrush’

The UK government has been advised by the House of Lords to adopt a more optimistic stance towards Large Language Models (LLMs) to avoid missing out on the AI boom. The report suggests focusing on near-term security and societal risks posed by LLMs, like copyright infringement and misinformation, rather than worrying excessively about hypothetical doomsday scenarios. It emphasizes the need for proportionate and practical regulation while avoiding regulatory capture by big tech companies like OpenAI and Google. The report also calls for enhanced governance measures to prevent regulatory capture and suggests addressing the use of copyrighted material for training LLMs promptly. TECHCRUNCH

Mamoon Hamid and Ilya Fushman of Kleiner Perkins: “More than 80%” of pitches now involve AI

Mamoon Hamid and Ilya Fushman, VCs at Kleiner Perkins, have been working to revitalize the firm. They've had success with investments like Rippling and Loom. They are now focusing on AI investments, with over 80% of pitches involving AI. They believe AI will revolutionize various industries, boosting productivity for high-paying jobs. They are also interested in AI engineers leaving big companies to start their own ventures. In terms of building moats, they look for strong customer engagement and a dominant market presence. Despite Figma's recent valuation drop, they remain optimistic about its future. Additionally, they are investing in AI wearables and exploring the next computing platform beyond mobile phones. TECHCRUNCH

AI is coming to the OnePlus 11 and 12 phones.

OnePlus is bringing AI features to its OnePlus 12 phones through a ColorOS update in China. This update includes AI-summarized phone calls and articles, as well as AI photo editing capabilities. While not as extensive as some other Android devices, OnePlus is showing a clear move towards integrating AI technology into its phones, primarily in the Chinese market. THE VERGE

AI helps reveal the ancient origin story of floral colors

Monash University researchers have used computer simulations to uncover the ancient connection between bees and the evolution of flower colors. They found that the first flowers evolved vibrant colors to stand out from their surroundings, attracting ancient pollinators like bees. By testing bees' color perception against simulated prehistoric environments in Australia, the study demonstrated that bees have UV, blue, and green photoreceptors, explaining why modern flowers often feature colors like yellow that are easily perceived by bees. This research could enhance our understanding of contemporary plant pollination and contribute to smart agriculture and efficient crop pollination techniques. PHYS ORG

A company lost $25.6 million because of a deepfaked conference call.

A company suffered a $25.6 million loss due to a fraudulent incident involving a deepfaked conference call. In this elaborate scheme, an employee was tricked into joining a video call with what seemed to be company executives, including the CFO, who were actually deepfake impersonations. The scammers on the call then instructed the employee to transfer the funds to five different Hong Kong bank accounts. This incident highlights the growing threat of deepfake technology in various forms of fraud, including cryptocurrency scams advertised on platforms like YouTube and even blackmail involving deepfake nude images. Vigilance and security measures are essential in combating these emerging threats. THE VERGE

The AI tools that might stop you getting hired

Employers are increasingly turning to artificial intelligence (AI) tools to streamline the recruitment process and monitor employees, but concerns are rising about their effectiveness and potential for discrimination. Researcher Hilke Schellmann conducted experiments with various AI hiring tools, revealing their flaws and biases. These tools include one-way video interviews, online CV screeners, game-based assessments, and social media analysis. Schellmann argues that these tools, often based on pseudoscience, may not be working as intended and can discriminate against certain candidates. She calls for more skepticism among HR departments, regulation, and transparency in the development of these AI tools to ensure fairness and effectiveness. THE GUARDIAN

Health experts raise concerns over AI apps claiming to be pocket nutrition assistants

Some health experts have concerns about AI apps claiming to be nutrition assistants. While AI can help with calorie counting and suggest meal options, human expertise remains important, especially for individuals with specific dietary needs or medical conditions. Dietitian-nutritionist Alexandra Corwin sees potential benefits in accurate AI food tracking but emphasizes the value of human nutritionists who can provide personalized guidance and motivation. Overall, AI can assist, but it can't replace the understanding and emotional support humans offer in nutrition management. FOX BUSINESS