Amazon Crashes the GAI Party with a Bang!
A few years ago, Amazon was hellbent on?transfer learning, a machine learning technique where pre-trained models are trained on a large dataset for a specific task and are used as a starting point for training a new model for a different, but related task. For instance,?Alexa?leverages?transfer learning?to learn new languages or tasks. But, sadly, the company?pulled the plug?on its voice-assisted feature a few months ago, succumbing to huge operating losses.
Unsurprisingly, Amazon is?not new to failures.?
The culture dates back to 2015, when Jeff Bezos, in a?letter to shareholders, wrote: “I believe we are the best place in the world to fail, and failure and invention are inseparable twins.” He later wrote in a?2018 letter?to shareholders: “As a company grows, everything needs to scale, including the size of your failed experiments. If the size of your failures isn’t growing, you’re not going to be inventing at a size that can actually move the needle.”?
Cut to 2023, with the buzz around Generative AI (GAI) – particularly large language models (LLMs) and multimodal models –?Amazon?has arrived?fashionably late?to the party. Last week, the company launched?Amazon Bedrock, alongside providing the general availability of?Amazon EC2 Inf2 instances?by AWS Inferentia2 chips; launching new?Trn1n instances?(powered by AWS Trainium chips); and offering free access to?Amazon CodeWhisperer?for individual developers.?
In February this year, AWS partnered with?Hugging Face?to accelerate the availability of next-generation machine-learning models alongside helping developers build, train, and deploy the latest ML models in the cloud using purpose-built tools.
Up until now, the GAI battle was largely between?Google?and?Microsoft. But now with AWS in the picture, the space is only going to intensify. One question remains –?Will AWS emerge victorious, or is this just a giant failed experiment in the making??
However, it is quite evident that cloud players hold an advantage due to their significant team size, robust infrastructure, extensive customer base, and easy accessibility. Clearly,?Amazon is ahead of the pack?– i.e. in the third quarter of last year, Amazon’s market share amounted to 34%, exceeding the combined market share of both?Microsoft Azure?and?Google Cloud.?
If we talk team size, too, Amazon wins. As per?Glass.ai, the company has the largest number of AI employees with a total of 10,113, followed by Microsoft (7,287, including?OpenAI),?Meta?(5,590), Google (5,341, including DeepMind) and?Apple?(4,773). It is also interesting to note that companies such as?OpenAI?(154),?DeepMind?(371),?Stability AI?(100), and?Midjourney?(11), which are spearheading AI breakthroughs, have relatively smaller AI teams.
Amazon chief?Andy Jassy?recently?told?CNBC’s ‘Squawk Box’ that most companies want to use large language models, but the really good ones take billions of dollars and years to train, and most companies do not want to go through that. He said that they want to work off of a foundational model that is “big and great already” and then have the ability to customise it for their own purpose. “That is what Bedrock is,” he added.?
Amazon Bedrock is a generative AI toolkit for chat, text and image. The platform has access to foundational models such as AI21 (Jurassic-2), Anthropic (Claude), Stability AI (Stable Diffusion) and Amazon (Titan).
Top Stories of the Week >>
The Crazy Food Habits of Tech Billionaires
“You are what you eat, you eat what you are, and you will definitely become what you put in your body, both from a moral and a physical standpoint.” This applies to everyone, including tech billionaires, who are at the pinnacle of success, and an inspiration to many.?
More than anything, their food habits give us a glimpse of the world around them and speak volumes about their leadership style and temperament. From Elon Musk to Sam Altman, Bill Gates, Mark Cuban and others, here is a quick look at their diet and what they like eating the most.?Read to find out.?
How AI is Exposing Our Dark Desires for the World’s End
“Everyone, deep in their hearts, is waiting for the end of the world to come,” is one of the most-popular quotes by Haruki Murakami in the novel ‘IQ84’, and also shown at the beginning of the movie ‘The Big Short’ (2015).
This statement can’t get any more real, especially with AI fueling the obsession like never before. For instance, recently, a user gave?ChaosGPT, a GPT-powered chatbot, the sole task to destroy humanity. The bot started scanning the web for the most-destructive weapons, and?said: “Human beings are among the most destructive and selfish creatures in existence.”?Read the full story here>>
Data Moats Have Fallen with GPT Clones
About five years ago, for a company to build a data moat was considered as important as storing your crown jewels in a bank vault. Even in the pre-internet era, companies like?IBM?had vendor data lock-in and specific details on the client’s business requirements.?
A strong data moat is why?Microsoft?bought?LinkedIn?in 2016 – it gave the software giant access to the data of more than 433 million members and a very competitive social graph. It’s also why?Twitter?cut the cord to its free API. But the irony is that with the passing of time and more innovations, the data land grab has come to mean nothing.?Intrigued? Read the full story here
AIM Videos >>
Meet Namma Yatri, the new player in the transport aggregation space that is trying to break the Ola-Uber duopoly in the country. The platform allows drivers to connect directly with customers, with zero commission fees. It is built on the foundation of the Beckn Protocol, backed by?Nandan Nilekani, chairman of Infosys.?
Watch the video to learn more about?Namma Yatri.
AIM Shots >>
In an?event?held at MIT,?Sam Altman?confirmed to Lex Friedman that?OpenAI?is?not working on GPT-5?anytime soon. He also addressed the letter on 6-months ban on?AI research and safety concerns.Read:?OpenAI Puts A Pause on GPT-5, Focuses on AI Safety.
Genpact and Dataiku?have?joined forces?to help organisations accelerate AI and ML initiatives to transform their businesses at scale.?Read more here.?
Amid the hype around?multimodal systems, Google researchers have developed a new model called?UniPi?that can learn how to do different tasks in different environments.?
In a bid to democratise LLMs,?Databricks recently released?Dolly 2.0. The unique aspect about this, and?other models, is that?it is available for commercial purposes without the need to pay for API access or share data with third parties.?Read more here.?
After?banning ChatGPT, the Italian government?is now looking to?reassess the situation. It has provided OpenAI with a set of regulations it needs to follow in order to resume services in the country.?Read more here.?
After a month?of launching?Consistency Models, OpenAI?has?open sourced?the model last week.?Find out more here.?
VideoVerse?recently?acquired?Reely.ai, an esports AI-powered?content creation, and social media distribution platform based in the US.?
Thank you so much for sharing this wonderful article with us. I believe, many people will find it as interesting as I do.