• Everyday AI
  • Posts
  • Everyday AI - "YOU WIN!" Microsoft Prevails

Everyday AI - "YOU WIN!" Microsoft Prevails

Two Big Announcements. One Clear Winner.

Everyday AI Logo

"YOU WIN!" Two Big Announcements. One Clear Winner.

Microsoft's fast moves forced Google to respond. The team at Redmond knocked it out of the park, while Google dropped the ball... twice.

The Big Stuff

Microsoft wins the first AI battle with solid launch event for Bing.

Microsoft presented their vision for search, a product category that has not changed materially in over 20 years. Microsoft's event was well-received, as they presented a vision for the future of search that included a chart similar to one Steve Jobs presented in 2007 at the iphone launch event. Microsoft's vision for the future of Generative AI feels like a similar technological shift to Apple's iconic device.

Microsoft/Apple Chart

Microsoft 2023 vs. Apple 2007 (iPhone Launch)

Microsoft showed that 50% of searches were going unanswered, and demonstrated how their GPT 3 technology could be integrated into search to provide more accurate and helpful results. They also unveiled a new feature in their Edge browser, which allows users to interact with search results in a sidebar.

SIGN UP HERE for the Bing waitlist!

I think this technology is going to reshape pretty much every software category

Satya Nadella, Microsoft CEO

Google. OOPs... and OOPs

Google's announcement of Bard seemed promising initially, but things went downholl quickly after that. They showed screenshots of ChatGPT competitor "Bard," but ultimately, they did not give a vision for search, which left people thinking, "Do they even have a strategy?" Even worse, the screenshot of bard had a factual error in it that got called out. Then, their live event in Paris was lackluster and felt defensive. At one point, they couldn't even find the demo phone to do a live demo on stage, and just bailed on the demo! As as a result of their missteps and Microsoft's strong showing, Google shed $150 BILLION in market cap this week.

At the event, Google effectively declared "you've been using our AI for years, but just didn't know it." They did not present a grand vision for the future. Although there were some impressive features for Google Maps, what people were looking for was how Google would make search more productive, which was not shown. 

Google did show some impressive computer vision products, such as a new version of image search, augmented reality overlays, and 3D representations of images in maps (NeRFs, or neural radiance fields). However, these products are not what Google makes their money on. It remains to be seen how they will respond to Microsoft's vision for search.

Let's Contrast that with Microsoft clear vision and dig a bit deeper in what they revealed:

Microsoft demonstrated four things in their new search vision:

  • OpenAI next gen large language model, optimized for search. More powerful than GPT-3, which powers ChatGPT

  • Prometheus model improves relevancy, annotates answers, provides up-to-date reulsts, and includes relevant geolocation

  • Core search index - largest jump in relevance in two decades

  • Unified UI/UX for the integrated product

Additionally, CEO Statya Nadella also said he was willing to accept any “de-monetisation” of the search business—from which Microsoft last year earned $11bn of revenue—for the chance to eat into Google’s business.Microsoft also demonstrated "Copilot" in their Edge browser, allowing users to summarize web pages, compare with other pages, rewrite code, plan trips, and generate social content. See a demo here. As a result of these announcements, Bing shot from 127 to number 6 in the app store.

It is becoming increasingly clear that Microsoft has been planning 2023 very carefully. They have been shipping at startup speeds, and have illustrated that the end game is integrating OpenAI into every part of their product, starting with Teams, then Bing, then the browser, and soon, Word, Powerpoint, and Outlook.

But is this really a paradigm shift, or is it just a fad? Time will tell, but ChatGPT had 672 million visitors in January (link)

Smaller, but still Cool Stuff

  • Microsoft deploys AI in the classroom to improve public speaking and math (link)

  • Researcher gets Bing search to reveal it's entire prompt, including it's codename, "Sydney" (link)

  • Reporting from @semafor claims OpenAI will release a ChatGPT mobile app soon. (link)

  • Want to work around ChatGPT's censorship? Just ask it to. (link)

  • Exclusive Q&A: John Carmack’s ‘Different Path’ to Artificial General Intelligence (link)

  • Google’s head of VR is leaving to form an AI company with the former co-CEO of Salesforce (link)

  • Google & HUJI Present Dreamix: The First Diffusion Model for General Video Editing (link)

  • David Guetta uses AI to create an Eminem sample. Crowd goes wild. (link)

  • Image to Sound demo - Take any image, and have a sound generated for it (link)

  • A Judge Just Used ChatGPT to Make a Court Decision (link)

  • The race of the AI labs heats up (link)

  • Stable attribution - find the artist behind the art (link)

  • TEXTure: Text-Guided Texturing of 3D Shapes w/demo (link) (link)

  • ChatGPT must be regulated and A.I. ‘can be used by bad actors,’ warns OpenAI’s CTO (link) (link)

  • BioGPT-Large, a GPT for medical text - Outperforms humans and it's going viral (link) (link)

  • FDA has now cleared more than 500 healthcare AI algorithms (link)

  • EchoNet-Peds is a model for assessing pediatric ultrasounds. It estimates ejection fraction comparable to human experts (link)

  • Seinfeld-inspired “Nothing forever” gets suspended from Twitch for anti-trans content (link)

  • Instructional image inpainting - allows you to give an instruction to edit images (link)

  • Claude, Anthropic AI's new chat is available via POE (link)

Incredible Resources: Just CLICK. Trust us.

AI has never been more accessible. Whether you’re just getting started, or you’re a veteran, there are some amazing resources to either help you get started, or take your abilities to the next level.

  • OASIS - Speak and let OASIS provide summaries for you. We created portions of this newsletter with it. Early access (link)

  • KnowledgeGPT - the power of GPT for your documents (link)

  • Kick Resume - Build your resume with the power of AI (link)

  • Genius - Copilot for Figma (link)

  • Galileo AI - Text to Design (link)

  • GPT Recipe Generator (link)

  • FinDataQuest - semantic search engine for financial data (early access) (link)

  • Job Description Generator (link)

  • Mokker.ai - Image Background Replacement (link)

  • Vowel AI - Meeting summaries (link)

  • Database co-pilot and client that can generate, run and iterate on queries from simple questions. Supporting SQL and MongoDB (link)

  • Roam Around - GPT travel advisor (link)

  • Fini - Turn your knowledge base into AI chat in 2 minutes (link)=

Get Your Nerd On: Whitepapers & Tutorials

  • Figures from Kosinski (2023). Theory of Mind May Have Spontaneously Emerged in Large Language Models (link)

  • Curious how diffusion models work? Read this thread. (link)

  • Toolformer: Language Models Can Teach Themselves to Use Tools (link)

  • Understanding Large Language Models -- A Transformative Reading List (link)

  • How to... use ChatGPT to boost your writing (link)

  • MIT researchers found that massive neural nets (e.g. large language models) are capable of storing and simulating other neural networks inside their hidden layers, which enables LLM to adapt to a new task without external training (link)

  • Vectorized object mapping of neural field slam - just watch the demo and it'll make sense (link)

Tweets of the Week

One Week is a Long Time in AI (oh believe us, we know) (link)

Largest Scale AI Computations doubling every six months (link)

Microsoft Copilot Creator, Alex Gravely, hints at a New Project (link)

ChatGPT was the 7th most popular page on Wikipedia in January (link)

ChatGPT Release was the Sputnik Moment oft

Eye Candy

NeRF (Look it up!) at the Met (link)

Make your own 2D cars using Midjourney (link)

Runway Gen-1 Demo (link)

Text-Driven 3D Texturing by Leonardo (link)

Chromium (link)

Memories from Travel (from @gliderceo)

Mark recreated this image from memories of travel in Amsterdam with his parents.

Color Out of Space (link)

What 45 MP looks like from Stable Diffusion (link

Pix 2 Pix Marble Terminator

ChatGPT - How Long Did it Take to Get to 1M Users?

Whew! Till next week.