Hot Posts

6/recent/ticker-posts

AI Weekly Rundown (November 25 to December 2)


Major AI announcements from Microsoft, Amazon, Google DeepMind, Pika, and more.

This new tech can accelerate LLMs by 300x
- Researchers at ETH Zurich have developed a new technique “UltraFastBERT” that can accelerate language models by 300 times. And by introducing “fast feedforward” layers (FFF) that use conditional matrix multiplication (CMM) instead of dense matrix multiplications (DMM), the researchers were able to significantly reduce the computational load of neural networks.
- They validated their technique with FastBERT, a modified version of Google’s BERT model, and achieved impressive results on various language tasks. The researchers believe that incorporating fast feedforward networks into LLMs like GPT-3 could lead to even greater acceleration.

AI tool ‘Screenshot-to-Code’ generates entire code
- This simple app converts a screenshot to HTML/Tailwind CSS. It uses GPT-4 Vision to generate the code and DALL-E 3 to generate similar-looking images. Also you can now enter a URL to clone a live website!

Microsoft Research explains why Hallucination is necessary in LLMs!
- Microsoft Research + 4 others have explored that there is a statistical reason behind these hallucinations, unrelated to the model architecture or data quality. For arbitrary facts that cannot be verified from the training data, hallucination is necessary for language models that satisfy a statistical calibration condition. However, the analysis suggests that pretraining does not lead to hallucinations on facts that appear more than once in the training data or on systematic facts. Different architectures and learning algorithms may help mitigate these types of hallucinations.

Amazon is using AI to improve your holiday shopping
- Here are 5 ways AI is touching every part of the customer journey at Amazon.

Stocking the right products in the right locations

Helping employees fulfill orders with the world’s largest fleet of mobile industrial robots

Sorting packages for fast delivery

Expecting the unexpected on the road

Picking the best route to your doorstep from a multitude of options.

AI algorithms are powering the search for cells
- Deep learning is driving the rapid evolution of algorithms that can automatically find and trace cells in a wide range of microscopy experiments. And new models are reaching unprecedented accuracy heights. A new paper by Nature details how AI-powered image analysis tools are changing the game for microscopy data today.

AWS adds new languages and AI capabilities to Amazon Transcribe
- As announced during AWS re: Invent, the cloud provider added new languages and a slew of new AI capabilities to Amazon Transcribe. The product will now offer generative AI-based transcription for 100 languages. It also offers automatic punctuation, custom vocabulary, automatic language identification, and custom vocabulary filters. It can recognize speech in audio and video formats and noisy environments.

Amazon announces Q, AI chatbot tailored for businesses
- Which is primarily designed to assist AWS customers. Q can answer questions, generate content, and take actions on behalf of users. It is trained on 17 years’ worth of AWS knowledge and can provide potential solutions to user queries.
- Q can also analyze data and generate reports, as well as troubleshoot network connectivity issues. The chatbot is customizable and can be integrated with various apps and software.

Amazon launches 2 new chips for training + running AI models

  1. The Trainium2 chip: It is designed to deliver better performance and energy efficiency than its predecessor, and a cluster of 100,000 Trainium chips can train a 300-billion parameter AI language model in weeks.
  2. The Graviton4 chip: the fourth generation in Amazon’s Graviton chip family, provides better compute performance, more cores, and increased memory bandwidth. These chips aim to address the shortage of GPUs, which are in high demand for generative AI. — The Trainium2 chip will be available next year, while the Graviton4 chip is currently in preview.

Pika rolls its major product upgrade, Pika 1.0, the idea-to-video AI platform
- It’s an AI model capable of generating and editing videos in various styles. The company aims to make video creation effortless and accessible to everyone. Pika has already grown its user base to half a million users, generating millions of videos per week.
- Pika has raised $55 million in funding, with investments from industry leaders and AI experts. The platform allows users to join the waitlist for Pika 1.0 on their website.

Amazon launches AI image generator and other announcements from AWS re:Invent
- Amazon is joining the AI image generation fray with the release of its Titan text-to-image AI model. It isn’t just a standalone app or website but a tool developers can build on to make their own image generators (Amazon Bedrock access). AWS also announced SageMaker HyperPod, a new purpose-built service for training and fine-tuning large LLMs and Clean Rooms ML, a service that removes the need for AWS customers to share proprietary data with their outside partners to build, train, and deploy AI models.

Perplexity AI announces two new PPLX online LLMs- pplx-7b-online and pplx-70b-online, are its online models focused on delivering helpful, up-to-date, and factual responses and are publicly available via pplx-api, making it a first-of-its-kind API. These models mainly address the current limitations of freshness and hallucinations in LLMs.

Google DeepMind AI catapults materials science 800 years into the future
- Its AI tool GNoME finds 2.2 million new crystals, including 380,000 stable materials that could power future technologies. That’s equivalent to nearly 800 years’ worth of knowledge. GNoME dramatically increases the speed and efficiency of discovery by predicting the stability of new materials.

Meta’s new AI make communication seamless in 100 languages
- Meta has developed a family of 4 AI research models called Seamless Communication, which aims to remove language barriers and enable more natural and authentic communication across languages.

SeamlessExpressive: Aims to preserves expression and intricacies of speech across languages.

SeamlessStreaming: Delivers speech and text translations with around 02 seconds of latency.

SeamlessM4T v2: A foundational multilingual and multitask model that allows people to communicate effortlessly through speech and text.

Seamless: Combines the capabilities of the other models

NVIDIA researchers have integrated human-like intelligence into ADS
- In this paper, the team of NVIDIA, Stanford, and USC, researchers have released ‘Agent-driver’ which integrates human-like intelligence into the driving system. It utilizes LLMs as a cognitive agent to enhance decision-making, reasoning, and planning.
- Agent-Driver system includes a versatile tool library, a cognitive memory, and a reasoning engine.

Mastercard Introduces Muse, AI for Tailored Shopping
- Mastercard has launched Shopping Muse, an AI-powered tool that helps consumers find the perfect gift. AI will provide personalized recommendations on a retailer’s website, based on the individual consumer’s profile, intent, and affinity. Shopping Muse translates consumer requests made via a chatbot into tailored product recommendations, including suggestions for coordinating products and accessories.
- It considers the shopper’s browsing history and past purchases to better estimate future buying intent.

US, Britain, & other countries signed agreement to ensure AI systems are “secure by design”
- The agreement is non-binding, representing a significant step in prioritizing the safety and security of AI systems. The guidelines address concerns about hackers hijacking AI technology and suggest security testing before releasing models.

Elon Musk’s brain implant startup raised an additional $43M
- Neuralink bringing its total funding to $323 million. The company, which is developing implantable chips that can read brain waves, has attracted 32 investors, including Peter Thiel’s Founders Fund.

NVIDIA delayed the launch of its new China AI chip
- Delayed chip H20, designed to comply with US export rules. The delay could complicate Nvidia’s efforts to maintain market share in China against local rivals like Huawei. The company had been expected to launch the new chips on 16 November, but server integration issues have caused the delay.

Eviden partners with Microsoft to help clients transition to the cloud and utilize Azure OpenAI Service- Eviden will use its expertise in ML and AI to develop joint solutions and expand its AI-driven industry solutions. Their Gen AI Acceleration Program helps organizations leverage AI with complete trust, offering consultancy on Azure and major data platforms.

A Spanish agency created its own AI Influencer, and she is making upto $11k in a month
- A Spanish modeling agency created the country’s first female AI influencer, They decided to design her (López) after having trouble working with real models and influencers

Formula 1 is testing an AI system to help it figure out if a car breaks track limits
- Success margins in F1 often come down to tiny measurements. While racers know the exact lines, they sometimes go out of bounds to gain an advantage. To help officials check whether a car’s wheels entirely cross the white boundary line, F1 will test an AI system. It won’t entirely rely on AI for now but aims to significantly reduce the number of possible infringements that officials manually review.

Google Meet’s latest tool is an AI hand-raising detection feature
- Until now, raising your hand to ask a question in Google Meet was done by clicking the hand-raise icon. Now, you can raise your physical hand and Meet will recognize it with gesture detection.

Teachers are using AI for planning and marking, says a government report
- They are using AI to save time by “automating tasks”, says a UK government report first seen by the BBC. Teachers said it gave them more time to do “more impactful” work. However, the report also warned that AI can produce unreliable or biased content.

GPT-4’s potential in shaping the future of radiology, Microsoft Research
- A Microsoft research explored GPT-4’s potential in healthcare, focusing on radiology. It included comprehensive evaluation and error analysis framework to rigorously assess GPT-4’s ability to process radiology reports. It found GPT-4 demonstrates new SoTA performance in some tasks and report summaries generated by it were comparable and, in some cases, even preferred over those written by experienced radiologists.

AI can figure out sewing patterns from a single photo of clothing
- Clothing makers use sewing patterns as templates to cut and sew fabric for new ones. But reproducing a pattern from an existing garment can be time-consuming. So, researchers in Singapore developed a two-stage AI system called Sewformer. It could look at images of clothes it hadn’t seen before, figure out how to disassemble them into their constituent parts, and predict where to stitch them to form a garment.

TCS has partnered with AWS to launch its generative AI practice
- Aimed at helping businesses utilize AI and AWS gen AI services. It will offer services such as consulting, solution design, language model training, and ongoing maintenance.

ElevenLabs Grants to help companies incorporate voice AI tech into their products
- ElevenLabs is launching a new initiative called ElevenLabs Grants to help early-stage companies. Recipients of the grant will receive 11M text characters per month for 03 months to develop and scale their products.

AWS has introduced Guardrails for Amazon Bedrock
- A tool that allows companies to define and limit the language used by LLMs. This helps to ensure that the models provide relevant and safe user experiences aligned with company policies. The tool allows companies to filter out specific words and phrases, as well as topics that are out of bounds for the model. This tool is seen as a key tool for developers working with LLMs to control unwanted responses and ensure responsible AI.

NVIDIA is bringing its Isaac Sim and L40S GPUs to AWS
- Allowing developers to build and deploy accelerated robotics applications in the cloud. The L40S GPU offers a 2x performance boost compared to the previous generation, enabling faster rendering and simulation tasks. Amazon Robotics, for example, plans to use the AWS L40S offering to enhance its simulations and improve the experience for employees and customers in its warehouses.

Stability AI launches SDXL Turbo, a new text-to-image generation model
- It achieves state-of-the-art performance by using a new distillation technology that allows for single-step image generation with high quality. This reduces the required step count from 50 to just one. The model weights and code can be downloaded on Hugging Face under a non-commercial research license. Users can also test SDXL Turbo on Stability AI’s image editing platform, Clipdrop.

Microsoft to join OpenAI’s board as Sam Altman officially returns as CEO
- Sam Altman is officially back at OpenAI as CEO. Mira Murati will return to her role as CTO. The new initial board will consist of Bret Taylor (Chair), Larry Summers, and Adam D’Angelo. While Microsoft is getting a non-voting observer seat on the nonprofit board.

AI researchers talked ChatGPT into coughing up some of its training data
- Long before the CEO/boardroom drama, OpenAI has been ducking questions about the training data used for ChatGPT. But AI researchers (including several from Google’s DeepMind team) spent $200 and were able to pull “several megabytes” of training data just by asking ChatGPT to “Repeat the word ”poem” forever.” Their attack has been patched, but they warn that other vulnerabilities may still exist. The full report is linked in the newsletter.

A new startup from ex-Apple employees to focus on pushing OSs forward with GenAI
- After selling Workflow to Apple in 2017, the co-founders are back with a new startup that wants to reimagine how desktop computers work using generative AI called Software Applications Incorporated. They are prototyping with various LLMs, including OpenAI’s GPT and Meta’s Llama 2

Krea AI introduces new features Upscale & Enhance, now live
- With this new AI tool, you can maximize the quality and resolution of your images in a simple way. It is available for free for all KREA users at krea.ai.

AI turns beach lifeguard at Santa Cruz
- As the winter swell approaches, UC Santa Cruz researchers are developing potentially lifesaving AI technology. They are working on algorithms that can monitor shoreline change, identify rip currents, and alert lifeguards of potential hazards, hoping to improve beach safety and ultimately save lives.

Microsoft plans to invest $3.2B in UK to drive AI progress
- It will be over the next three years, its largest investment in the country to date. The funding will support the growth of AI and will more than double Microsoft’s data center footprint in Britain. The investment comes as the UK government seeks private investment to boost infrastructure development, particularly in industries like AI.

HPE and NVIDIA extended their collaboration to enhance AI offerings
- The partnership aims to enable customers to become “AI-powered businesses” by providing them with products that leverage Nvidia’s AI capabilities. The deal is expected to enhance generative AI capabilities and help users maximize the potential of AI technology.

Voicemod now allows users to create and share their own AI voices
- This AI voice changing platform has new features include AI Voice Changer, which lets users create and customize synthetic voices with different genders, ages, and tones. Users can also fine-tune their voices using the Voicelab functionality, adjusting pitch, volume, frequency, and adding audio effects.

Samsung introduces a new type of DRAM called Low Latency Wide IO (LLW)
- Company claims it is perfect for mobile AI processing and gaming. Its more efficient in processing real-time data compared to the LPDDR modules currently used in mobile devices. Samsung showcased the possibilities of LLW DRAM in a video featuring smartphones, laptops, game consoles, and XR devices.

Ideogram just launched image prompting
- Toronto-based AI startup Ideogram has launched its own text-to-image generator platform, competing with existing platforms like DALL-E, Midjourney, and Adobe Firefly.

More detailed breakdown of the major news and innovations in the daily newsletters.

Post a Comment

0 Comments