IBM announced the release of its new AI and data platform, IBM watsonx, at its recent annual Think conference. The platform aims to help companies scale and accelerate their AI with trusted data, providing everything from hardware to data storage and ML and AI foundation models. IBM’s new AI products are a direct response to others like OpenAI’s chatbot ChatGPT, which has been dominating public awareness since December.
It may all be a little too late for IBM to take the lead in the market.
They're not the first ones to come out with this sort of platform,” says Bradley Shimmin chief analyst for Omdia said in an interview with Data Center Knowledge. “Google and AWS’ Databricks, pretty much any major platform that has AI development tooling has already stepped up to work on this. I don't think IBM is in a position to showcase leadership unless they, on the research side of things or surprise us with something that looks more like the PaLM 2.0 announcement from Google earlier last week.”
What Are the Components of IBM's watsonx Platform?
The watsonx platform includes two parts: watsonx.ai and watsonx.data. Watsonx.ai is an AI studio that combines IBM Watson Studio's capabilities with the latest generative AI capabilities that leverage the power of foundation models. IBM’s watsonx.data is a fit-for-purpose data store built on open lakehouse architecture optimized for governed data and AI workloads.
It is expected to be generally available by July 2023 and can manage workloads both on-premise and across multi-cloud environments. The solution will also provide built-in governance tools, automation, and integrations with an organization's existing databases and tools to simplify setup and user experience.
To shift customers from being just "users," IBM is also creating a governance toolkit as part of the watsonx offering.
"They can become AI advantaged,” said Arvind Krishna, IBM Chairman and CEO in a statement. "Foundation models make deploying AI…scalable, affordable, and efficient. With IBM watsonx, clients can quickly train and deploy custom AI capabilities across their entire business, all while retaining full control of their data."
Additionally, the company promises a reduction in data warehouse costs by as much as 50%.
Shift in Focus from Gigantic AI Models to Effective Solutions
While IBM's watsonx platform aims to be a leader in the AI space, Sam Altman, CEO of OpenAI, believes that the era of giant AI models may be coming to an end. At an MIT event in April, Altman declared that future progress in AI needs new ideas, not just bigger models.
“I think we're at the end of the era where it's going to be these, like, giant, giant models,” Altman told the audience. “We'll make them better in other ways.”
However, it is worth noting that watsonx may be behind some of the AI innovations that Altman is referring to. With its machine learning and AI foundation models, watsonx is enabling enterprises to quickly train and deploy custom AI capabilities, which could lead to new breakthroughs in the field.
“While ChatGPT can write you jokes, pictures and even entire songs and movie scripts, these type of consumer AI outputs are not what businesses need,” said Tarun Chopra, vice president of IBM product management, data and AI said in an email. “More than ever, businesses just starting on their AI journeys need seamless introductions to the technology.”
Businesses, Chopra says, need absolute confidence that the AI being used for mission-critical decisions and outputs are trustworthy and reliable while working off their proprietary data.
Additionally, “watsonx allows businesses to own their data on their own infrastructure,” Chopra said.
Shimmin says he wouldn’t compare the two platforms as OpenAI is building a closed system with a number of emergent properties that allow users to do many different modalities with one model.
“What IBM is talking about there is helping enterprises to reach the same type of functionality that you get from a GPT model but to do so in a, in a fashion that the enterprise is more accustomed to and more comfortable with,” explained Shimmin.
OpenAI’s been able to scale a series of consumer-ready, advanced AI by taking existing machine-learning algorithms and scaling them up to previously unimagined size. GPT-4, the latest of those projects, was likely trained using trillions of words of text and many thousands of powerful computer chips. The product's rapid deployment has triggered an AI arms race in the tech industry and raised alarm among some AI ethicists and public officials, who fear the technology's potential to spread misinformation, replace jobs or otherwise cause significant harm to users.
This is something watsonx would eliminate added Chopra as IBM has focused on “meticulously curating everything that goes into our models.” The foundational model was developed to folder data for hate, profanity, licensing restriction, and bias.
However, Shimmin says Speaker OpenAI is working towards allowing companies to bring their own data in a safe manner. Salesforce, Shimmin says, partners with OpenAI to create a “walled garden” that is their own iteration of OpenAI to control and secure data access.
IBM Watson AI, the Class in Overpromising and Underdelivering
Watson gained fame when it competed on the popular network game show "Jeopardy" against two of its most successful contestants, Ken Jennings and Brad Rutten. As anticipated, Watson crushed its human competitors winning $77,147 compared to $24,000 and $21,600 for Jennings and Rutter, respectively.
Watson’s fame was short-lived and seen as a one-trick pony while becoming a cautionary tale of tech overpromising and underdelivering. But a newly energized AI trend -- prompted by OpenAI’s ChatGPT-4 could spark a renewed interest in the public’s awareness of Watson AI to the public. This could make up for the rise and fall of Watson AI in 2011 which left some concerned about IBM's commitment to emerging technology. It’s one of the reasons, IBM enabled Watson and its suite of AI tools to run in the data center of its biggest cloud infrastructure rivals: Amazon Web Service and Microsoft Azure. Operating software that trains a machine learning model is cheaper and more efficient when done in the same location.
Additionally, IBM’s first Watson didn’t catch on as it was costly for companies to adopt, according to previous reporting from Reuters. IBM is now marketing the business-focused watsonx as an A.I. development platform that businesses could use to build their own models for a number of things, from customer care or writing code.
There’s been a scramble, Shimmin says, to innovate on the smaller, meta-based models, such as Facebook’s LLaMA based models that drive enterprises.
Shimmin says it isn't just the handful of super large language models like GPT and PaLM, but instead, smaller models that are tuned with corporate data and that they'refine tuning and is done in a secure, governed, managed, transparent environment.
“And that's what IBM is setting up here with watsonx,” says Shimmin, “so I applaud them for the approach.”
Will watsonx Lead the Enterprise AI Solutions Race?
As the AI industry continues to rapidly expand, IBM's one-stop-shop approach to AI and server needs may give it an edge over competitors. While AI investments continue to trigger talks of a global shift similar to the Industrial Revolution, many, including IBM’s own workers, are concerned at the thought of the company’s commitment to the emerging technology, given as it was reported earlier this month that IBM would be pausing hiring in favor of AI, with as many as 7,800 jobs at risk.
In a Q&A ahead of IBM’s Think conference, Krishna predicted that A will take over “repetitive back-office processes” from human workers.
“We see this easily taking anywhere from 30 to 50% of that volume of tasks and being able to do them with really as much or better proficiency than even people can do,” Krishna said.
“That lot, we see getting embraced right away starting this year, and getting to full fruition over the next three to five years,” added Krishna.
