Google's AI Avalanche, a $6B AI request from the Airforce, Censorship AI, & AI gives speech to Paralysis Patient
The A.I. Executive Briefing is an expert weekly curation of A.I. news by our research team, shared externally now because we feel there’s too much hype & noise in the market. The same content will be distributed through?this substack .
NEWS ROUND-UP
Product
1. Google AI Avalanche
2. Hugging Face Cloud Computing
3. AmazonOne: Palm ID
4. Poe’s new app is a Chatbot LLM Aggregator
5. Content Moderation with AI
Regulation
6. America Aims at AI
Research
7. AI nose and AI ears
Healthcare
8. AI Enables Speech for a Paralyzed Woman & Stanford Scrapes Medical Twitter
VENTURE NEWS
9. Imbue raises $200M to build AI models that can ‘robustly reason’
10. Generative AI startup AI21 Labs raises Series C for $155M at a $1.4B valuation
11. US AI startup Poolside raises $126m seed round and relocates to France
12. AI chip startup d-Matrix raises $110 million in Series B
13. Robot software firm Mujin raises $85M Series C
14. ThetaRay raises $57M series D to build AI tools that fight money laundering
15. Rockset, real-time database for AI, raises Series B of $44M
16. Intuition Robotics raises $25 million to tackle loneliness with AI
17. Harmonya raises $20 million Series A to assist retailers with AI-powered product data
18. Speak, backed by OpenAI, Announces $16m Series B-2 Financing
NEWS ROUND-UP
Product
1. Google AI Avalanche
谷歌 has launched a myriad of new AI products and features after their NEXT conference . Their long awaited AI assistant, Duet AI , will integrate into drive, docs, sheets, slides and other products in that suite. Duet AI is a real-time AI assistant that can do things like write documents, create presentations, and even attend meetings on your behalf. Deepmind has also released a new Watermarking system in images?generated by Imagen with Deepmind’s SynthID will help humans distinguish generative deep fake images.
In terms of partnerships, Fox will use Google’s Vertex AI Vision to query their database of 1.9M sports videos to produce new content in real time. GM introduced Onstar Virtual Assitant in 2022. It used Google AI tech to respond to FAQs and navigate. They expanded the partnship with Dialogflow upping capabilities to respond to 1M+ inquiries. It will also listen to your convorsations for “emergency” words
BigCommerce , an e-commerce platform that allows users to build and host online stores, unveiled its partnership with Google Cloud to add AI-powered features to its open SaaS eCommerce platform. BigCommerce plans to leverage AI capabilities, particularly?Google’s vertex API , to enhance and simplify the task of product descriptions.
It is also worth noting that Google’s Cloud service adds 20 new prebuilt AI models optimized for enterprises, with customers gaining access to third-party systems like?LLaMa-2 ?(Meta) and?Claude 2 ?(Anthropic).
Why it matters: Around the same time that google announced the their new Duet AI, Zoom has also launched their own meeting assistant that can largely do most of the tasks duet is assigned to do (for google meets). Although more people in inevitably use Duet AI over Zooms AI assistant, the combined distribution of their meeting and task assistance will help shift public sentiment around AI over time. Unexpectedly, with their BigCommerce partnership, it seems Google is directly challenging Shopify & Amazon. It is important to note that they are powering back end tasks with their Vertex technology that Fox is also leveraging. The fact that they are stepping into the ecommerce arena specifically with their AI tech is a noteworthy action. Google Cloud service is also taking a step towards building a platform for AI builders to come and leverage multiple models. We saw Amazon take a step in this direction with Amazon Bedrock & we also have news of Poe’s new desktop app also have a similar ability (see below). It seems that google is capable of competing at the toes of major players in multiple industries when it comes to bringing in new AI technology.
2. Hugging Face Cloud Computing
The open-source repository's 'Training Cluster as a Service' gives users access to Hugging Face ’s own GPUs – of which the company has thousands, including?Nvidia H100s ?and?A100s . With all the new money that Hugging Face has been raising this year, there were questions about how an open-source first project would eventually make enough money to pay back their investors.
Why it matters: Hugging Face is one of the most well-capitalized and renowned AI start-ups in the space right now. They can continue to be one of the “good guys” of the space by democratizing access to increasingly rare HPC NVIDIA chips. Luckily for them and their cap table, the company will also be able to profit from this initiative. As long as they can logistically maximize their GPU usage, anyone who is building AI models on their platform and needs access to extra compute power can now acquire it within their ecosystem. Hugging Face is clearly looking at taking over the whole AI/ML app development platform from end to end.
3. AmazonOne: Palm ID
This new technology is a combination of AI, biometrics, and computer vision. The system scans your hand to create your palm signature and connects it to your credit card or Amazon account. The neural network has been trained on millions of synthetically generated images of the palm and the vessels underneath allowing us to boost the system’s accuracy. Amazon One has already been used more than 3 million times with 99.9999% accuracy. Amazon One is currently being released across 500+ Whole Foods Markets and dozens of stores, sports venues, grocers, convenience stores, and entertainment venues.
Why it matters: New methods of biometric scanning and Identification are the natural response to a world where generating digital identities comes with ease. Few concepts seem more secure than proving your own identity with your own body. What is particularly interesting is how Amazon trained their model, which is with synthetic data! We have covered the dangers of using synthetic data for training language models— as over time, it will degrade the quality of outputs. However, for Computer Vision training data, as long as biases are considered when selecting/generating training data, synthetic data works well. This fact has been proven out by Autonomous Driving companies utilizing synthetic geographical data to train driving models.
4. Poe’s new app is a Chatbot LLM Aggregator
Poe, the?AI chatbot platform ?created by Quora , has added a slew of updates, including a Mac app, the ability to have multiple simultaneous conversations with the same AI bot, access to Meta’s Llama 2 model, and more. It’s also planning an enterprise tier so that companies can manage the platform for their employees, according to an email that was recently sent to Poe users. “I think there’s going to be this massive ecosystem similar to what the web is today,” Quora CEO Adam D’Angelo?recently said . “I could imagine a world in which most companies have a bot that they provide to the public.” Poe lets you pay one subscription for unlimited access to all of the bots on its platform for $19.99 per month or $200 per year.
Why it matters: Andrew: According to Adam D’Angelo, the CEO of Quora (who also sits on the board of OpenAI) Poe’s ambition is to be the web browser for AI chatbots. In my opinion, I think his insights are correct but his conclusion is completely wrong. I don’t believe we’ll have this arcade-type of destination like Poe where you can select from hundreds or thousands of A.I. bots. I think what Quora & Poe should focus on is similar to what StackOverflow is doing in terms of leveraging A.I. and their community (which is incredibly valuable and difficult to replicate) to deliver a next-gen user experience.
领英推荐
5. Content Moderation with AI
Activision said that they are partnering with AI-based Modulate’s ToxMod voice chat moderation system which will debut with the release of Call of Duty: Modern Warfare III on November 10. It’s a big win for Modulate and its ToxMod technology, which identifies toxic speech, including hate speech, discriminatory language, and harassment in real-time — and then automatically deals out the consequences. This system will complement Call of Duty’s existing moderation efforts, which include text-based filtering across 14 languages and a robust in-game reporting system. On a similar note (but a completely different industry), dating app company Bumble has released its?updated community guidelines ?with an aim to crack down on bots, spam, ghosting and doxing. The new policy will apply to its dating app Badoo along with Bumble. The platform is revamping its rules to weed out bots by banning actions like artificially influencing “connections, matching, conversations, or engagement through the use of automation or scripting.” Bumble specified that it has blocked over 8.2 million accounts this year across its properties — a majority of them through the company’s AI to detect fake and spam accounts.
Why it matters: Andrew: During my time in venture capital, I saw over 100 decks/companies that tried to use ML for content moderation across different industries (gaming was a big one). It’s great to see this technology being more widely deployed across major platforms, across different industries, and huge kudos to startups like Modulate to supply this valuable technology to major juggernauts like Activision.
Regulation
6. America Aims at AI
As the American public continues to fret about how their personal data will be used in Google or Meta’s next AI product, the US Government is more concerned with what countries are getting access to the latest chips being released. In NVIDIA’s latest filing they stated, "During the second quarter of fiscal year 2024, the U.S. government informed us of an additional licensing requirement for a subset of A100 and H100 products destined to certain customers and other regions, including some countries in the Middle East.” A similar notice was also shared with AMD.
America will also be hosting another AI Summit in its continued effort to construct public facing law regarding AI systems and technologies. The September 13th event will involve Google CEO Sundar Pichai and former Google CEO Eric Schmidt; Meta CEO Mark Zuckerberg, OpenAI CEO Sam Altman; Microsoft CEO Satya Nadella; Nvidia CEO Jensen Huang; and Elon Musk, CEO of X.
Interest in using AI (rather than regulating it) is also rising in the government. The Copyright Office is undertaking a study on AI and copyright issues, particularly relating to generative AI that can produce text, images, audio, and video. The study will focus on several key topics:
Agencies such as the National Archive want to use AI to help keep its inventory of documents, a very applicable and tame use case that could make a big impact in document archival and retrieval. Other agencies, such as the Airforce, have requested?nearly $6 billion in federal funding ?over the next five years to construct a fleet of XQ-58A Valkyrie uncrewed aircraft, each of which will cost $3M.
Why it matters: The US Government is in an interesting place where agencies are simultaneously exploring, regulating & utilizing AI systems. Defense & Warfare is the most expensive line item in the government’s budget, so it only makes sense that the Airforce would ask for $6B in order to build new AI powered drones. The national archive utilizing AI to construct more accurate meta data for their archived books increases effeciency and is a very accessible use case of AI systems. The fear of powerful AI systems being built in China expanding to fear of the Middle East is surprising although likely related to the global shift in political power showing signs of skewing away from the West with new Associations such as BRICS+ and OPEC+. The US is going to attempt to keep control of high tech industry while it seems many other emerging nations are more focused on natural resources and currency control.
Research
7. AI nose and AI ears
You’ve heard of LLMs imitating human speech and you’ve heard of Computer Vision mimicking human sight, but how about AI-enabled noses and ears?
A research team co-led by the?Monell Chemical Senses Center?and start-up?Osmo, which spun out of Google Research, is investigating how airborne chemicals connect to odor perception in the brain. Based on molecular structures, researchers have developed an AI system that forecasts the aroma of compounds. Human panelists trained in the study when compared to the AI model, found that the model outperformed individual human assessments for 53% of the molecules tested.
In other news and AI AI-powered hearing aid is offering more than just breaking the silence for the hearing impaired. Starkey's Genesis AI hearing aids can make what you hear crisper, but they can also serve as a personal assistant. You can use them as a phone, or listen to music. There's an accidental fall detection feature and foreign language translation.
Why it matters: Smell is one of the most complex sensory abilities that humans have. It has an unique association with memory and accesses different parts of the brain than hearing and sight do. We have observed the application of molecular structure prediction to drug discovery, but this research is the first instance of using molecular structures to predict their effects/smell. The application of such technology will be extrapolated to any form of molecule detection, which can easily be installed as a small device in public spaces in order to detect any dangerous chemical compounds. The advancement in hearing aid technology is a step in the direction where we can improve senses for the sensory impaired beyond that of typical human abilities. Installing a hearing aid with real-time language translation is functionally as good as learning that language, Imagine learning 80 languages at once! It's like Neo downloading Kung Fu from the Matrix!
Healthcare
8. AI enables speech for a Paralyzed Woman amp; Stanford Scrapes Medical Twitter
UCSF researchers have developed an implantable AI-powered device that, for the first time,?translates brain signals into modulated speech and facial expressions . Thanks to this groundbreaking medical device, a woman who lost the ability to speak due to a stroke was able to speak and convey emotion using a talking digital avatar.
Additionally, Stanford researchers scrape medical Twitter (X.com ) for more than 200,000 anonymized diagnostic images medical images, and diagnosis information of cancers and other diseases, along with the professional annotations provided by those highly trained clinicians. Utilizing this curated dataset of medical images, diagnoses, and annotations, the model has been able to correctly analyze and diagnose previously unseen images.
Why it matters: First and foremost, the ability to express brain signals into digital representations of the intended action is genuine sci-fi and is a remarkable medical breakthrough for paralyzed people. The next step is to link this to one of the many AI-powered robotics companies and we are one step closer to transferring consciousness! The news of Stanford researchers successfully scrapping over 200k annotated medical images was surprising given all of the drama surrounding their API and the crisis other forums were in over the “illegal” training that occurred with OpenAI, PaLM, and LLaMA LLM models. Given the circumstances, and the quality of the data they extracted from the platform, I would say this is a success for the open-source movement and for the healthcare community overall.
VENTURE NEWS
9. Imbue raises $200M to build AI models that can ‘robustly reason’
Imbue raised ?$200M in a Series B funding round that values the company at over $1 billion. Among those participating are the Astera Institute, Nvidia, Cruise CEO Kyle Vogt, and Notion co-founder Simon Last. Imbue is developing proprietary models that are “internally useful” to start, including models that can code. These models with over 100B parameters are optimized to perform well on internal benchmarks for reasoning. This training is being conducted on a compute cluster co-designed by Nvidia, containing 10,000 GPUs from Nvidia’s H100 series, A common standard we have noticed in many successful AI companies.
10. Generative AI startup AI21 Labs raises Series C for $155M at a $1.4B valuation
AI21 Labs , which provides text-based generative AI services to enterprises, has closed a $155M series C funding round. This round values the company at $1.4B. In 2020, AI21 Labs launched?Wordtune , an AI-based “writing companion that understands context and meaning.” It rose to popularity in 2021 and now that the ChatGPT mania has lightened up, AI21’s proprietary LLM, Jurassic-2, has also been gaining credibility. This is the model that powers Wordtune & AI21 Studio, which is a developer platform for building custom text-based business applications.
11. US AI startup Poolside raises $126m seed round and relocates to France
Poolside has raised from French billionaire Xavier Niel and US VC Felicis led the round of $100M as an extension to a $26M seed round announced in May, which was led by US-based Redpoint Ventures. Poolside was founded in May 2022 by Jason Warner, former CTO of GitHub, and Eiso Kant, a serial entrepreneur. Poolside’s Paris funding party comes a few months after French president Emmanuel Macron announced a €500M plan to encourage the development of new AI players, sending a strong signal that the country is open for business.
12. AI chip startup d-Matrix raises $110 million in Series B
The Series B funding round was led by the Singapore-based Temasek, and included the Palo Alto, California venture firm Playground Global and Microsoft. d-Matrix designs chips that have digital "in-memory compute" that enables AI computer code to run more efficiently. The company's chip technology uses less energy to crunch data required to spit out generative AI responses and is optimized for such tasks. D-Matrix projects under $10M in revenue this year - largely from customers purchasing chips to evaluate. The business expects to generate more than $70M to $75M in annual revenue in two years.
13. Robot software firm Mujin raises $85M Series C
Mujin announces its $85M Series C, bringing its total funding up to $150M. SBI Investment led the latest round, with participation from Accenture. The company develops AI-based software for industrial automation and believes that robots can improve productivity and safety. Its platform is designed for picking and moving e-commerce product on and off of pallets. Earlier this year, the company unveiled a truck bot that can unload trailers and shipping containers.
14. ThetaRay raises $57M series D to build AI tools that fight money laundering
ThetaRay raised $57M in a private funding round led by global fintech platform Portage. The company has raised around $160 million to date — with previous backers including strategics like ABN Amro, as well as General Electric, Alibaba, PwC, and SVB Financial. The company is building an anti-money laundering (AML) platform that automatically scans and identifies illicit transaction activity at banks and other financial institutions.
15. Rockset, a real-time database for AI, raises Series B of $44M
Rockset raises a $44M series B round led by?Icon Ventures , with participation from?Glynn Capital ,?Four Rivers ,?K5 Global ,?Sequoia ?and?Greylock . The company is innovating on the RocksDB framework first developed at Meta and has recently adapted its real-time indexing technology to work for vector databases.
16. Intuition Robotics raises $25 million to tackle loneliness with AI
Intuition Robotics raises a $25 million funding round, of which $5M is venture debt. The round was led by Woven Capital and the growth fund of Toyota. Intuition Robotics’ ElliQ is an AI companion designed to improve the lives of the elderly by combating loneliness and promoting healthy and independent lives. The ElliQ companion tool has already been deployed in the state of New York and has shown results demonstrating that users engage with ElliQ up to 30 times per day, reducing loneliness for 95% of its users.
17. Harmonya raises $20 million Series A to assist retailers with AI-powered product data
Harmonya , announced a $20 million Series A funding round. The round was led by Bright Pixel Capital with follow-on from Team8 and Arc Investors. The company offers an AI-powered product data enrichment, categorization, and insights platform for retailers and brands. They specifically leverage LLMs to dynamically surface product attributes from a wide range of sources.
18. Speak, backed by OpenAI, Announces $16m Series B-2 Financing
Speak has raised $16M in a Series B extension round. The round was led by Lachy Groom, with participation from the OpenAI Startup Fund and new strategic investors such as the founders of Dropbox, Drew Houston, and Arash Ferdowsi. Surprisingly, the company has captured “6% of Korea’s population” who are interested in learning English. The team exclusively focuses on solving the Korean population's need to learn English via their two-way dialogue-based platform.
Send us a message with any questions/comments/thoughts on anything A.I. related and we’ll try to answer them in our next release.
Expertise in Text Analytics | Conversational and Customer Feedback Platform
1 年Love the image
Managing Partner at Geek Ventures. Two 100x exits. Investor in early stage tech companies that change the world. Seed investor in a Decacorn.
1 年great article!