But just like the cryptocurrency gold rush, scammers aren’t the only ones rushing to take advantage of the new generation of machine learning software, so-called “generative AI” software. The software can perform a variety of tasks from writing sentences to writing poetry to painting. From photography to creating software to filling out forms.
The world’s biggest software companies, including Google’s parent company Alphabet, Microsoft, China’s Baidu, Salesforce and even Meta itself, are flocking to the space opened by OpenAI’s ChatGPT.
And the researchers and executives of those companies have apparently retired en masse and founded their own machine learning startups, often working on niche applications of generative AI.
Power your search engine
In February, just 68 days after OpenAI announced ChatGPT, Alphabet’s two leading AI companies, Google AI and DeepMind, announced that they would offer many of ChatGPT’s features, but their own machine-based Announcing Bard, a generative AI service. -Language Model for Dialogue Applications (or LaMDA for short), a learning model that Alphabet has been working on for many years.
The next day, Microsoft embarked on the move, equipping its struggling search engine, Bing, with machine learning software from OpenAI, which Microsoft has backed since 2021, and injecting another $14 billion in January. announced that it would replenish its investment.
“AI will fundamentally change every software category, starting with the largest of them: search,” said Satya Nadella, Microsoft chairman and CEO at the announcement of Bing. said.
Microsoft CEO Satya Nadella is stepping up investments in artificial intelligence as competition in generative AI heats up. bloomberg
And it didn’t take long for that prediction to start becoming a reality.
In March, Chinese search engine giant Baidu also announced it would be launching a ChatGPT rival known as “Weixin” (or “Ernie”), whose bot will go beyond simply finding answers to questions to solve complex math problems. You can also explain step-by-step how you arrived at that answer.
But, as Nadella suggested, using generative AI to power searches is just the beginning of the applications this technology can be applied to.
Earlier this month, customer relationship management (CRM) software giant Salesforce also flocked to the generative AI space, collaborating with OpenAI on Einstein GPT, a chatbot that allows Salesforce customers to send personalized emails to their customers. announced. Respond to customer inquiries and help software developers write code.
serious change
“The world is undergoing one of the most profound technological changes with the rise of real-time technology and generative AI,” Salesforce CEO Marc Benioff said at the launch.
Importantly, Einstein ingests vast amounts of public data from the internet to create so-called “large scale language models” that can accurately guess the words that make up the answer to an incoming question, machine learning created by OpenAI. It wasn’t just using the model. From a ChatGPT user.
Salesforce Founder and CEO Marc Benioff Oscar Coleman
Einstein users can also use their own private data to retrain machine learning models in real time, ensuring that Einstein-generated content is relevant and up-to-date, Salesforce says.
Meta’s own work on large language models is designed to help with such retraining of machine learning models as well.
In February, the company announced a large language model known as LLaMA (Large Language Model Meta AI). It comes in a variety of sizes, allowing researchers to choose the most cost-effective size for retraining.
“Despite recent advances in large-scale language models, full research access to them is still limited due to the resources required to train and run such large-scale models,” Meta said. said Mr.
“Smaller models trained with more tokens (parts of words) are easier to retrain and fine-tune for specific potential product use cases. , trained LLaMA 65B and LLaMA 33B with 1.4 trillion tokens.Our smallest model, LLaMA 7B, has been trained with 1 trillion tokens.”
The high cost of training new machine learning models is the main dividing factor between the ChatGPT boom and the cryptocurrency gold rush that preceded it.
While startups have been able to create their own cryptocurrencies at low cost, with some estimates producing over 23,000 different cryptocurrencies, creating new large-scale language models requires access to large amounts of data and , requires a large amount of computer power to process that data. It can turn out to be expensive.
That has been a barrier to entry for startups, and in the months since ChatGPT sparked global interest in generative AI, startups are well-funded and already have researchers and entrepreneurs with a proven track record in the technology. tend to be established by
In May, the two Google researchers who sparked the generative AI boom with their 2017 paper, “Attention is All You Need,” raised $8 million in a funding round led by New York’s Thrive Capital. reportedly received $12.5 million. A venture capital firm that also backs OpenAI.
The two researchers, Ashish Bhaswani and Niki Palma, are actually leaving Google in 2021 to found another machine learning company, Adept AI, which also owns machine learning for enterprise customers. It has raised $415 million to develop the application.
Their latest startup, known as Essential AI, was still in stealth mode at the time of writing, but reports say it’s not as enterprise-focused as Adept AI.
They’re not the only entrepreneurs to leave big AI companies to found startups.
Noam Shajeer, listed as the second author (after Ashish Bhaswani and just before Niki Palma) in Google’s influential 2017 research paper, founded Character.ai. The startup has raised about $200 million to build a chatbot that mimics a person’s personality. A person who is famous or has a particular profession.
Meanwhile, Aidan Gomez, listed as the sixth author of “Attention Is All You Need,” recently revealed a similar to Salesforce’s Einstein chatbot that companies can deploy to streamline processes. Raised $250 million for Cohere AI, his startup that builds tools.
(In fact, Salesforce is reportedly one of the backers of Cohere’s latest funding round, possibly drawing from the $250 million Generative AI Fund Salesforce announced when it revealed Einstein. presumably procured).
The investment in generative AI is huge, with market research firm IMARC Group estimating that the global AI market is now worth $70.9 billion, growing at a compound annual rate of 30%.
This will result in a value of $350 billion by 2028, IMARC Group estimates.
