In late 2022 or early 2023, the emergence of consumer-facing AI tools radically changed public perceptions of AI's power and potential. Although generative AI has been a hot topic among experts ever since GPT-2 was introduced in 2019, its potential for revolutionizing enterprise is only now becoming clear. This moment, and the ripple effect it will have on future generations, is a weighty one.
The impact of generative AI will be revolutionizing on the economy and business. McKinsey Global Institute estimated that generative AI would add between $2.6 trillion and $4.4 billion in value annually to the global economies, increasing the impact of AI on the economy as a entire by 15 to 40 percent. The consultancy estimates that AI will automate 50% of all work by 2040-2060. Generative AI is expected to push that date a decade sooner than previous predictions. Goldman Sachs projects a 7.5% increase in global GDP, or nearly $7 trillion, due to generative AI. The firm also expects two-thirds U.S. occupations to be affected by AI powered automation.
Large language models (LLMs) are the basis of text-generating AI systems such as ChatGPT. LLMs are trained on vast data sets to answer questions and perform tasks using statistical probabilities. Instead of searching for answers and synthesizing them, they use mathematical modeling to predict the next most likely word or output. Michael Carbin is an associate professor at MIT, and the founder of MosaicML. He says, "What I found exciting was how conversant ChatGPT was when I first interacted." "I felt, for the very first time, that I could communicate with a machine and it would understand what I was saying," says Michael Carbin, associate professor at MIT and founding advisor at MosaicML. Now we can translate language so that machines can understand it. "I can't imagine anything more powerful than the desktop computer."
Our survey of CIOs in 2022 found that, although AI was recognized to be strategically important, CIOs had limited ambitions. While 94% organizations used AI, only 14% aimed to achieve "enterprise wide" AI by the year 2025. The power of generative AI to democratize AI – to spread it throughout the enterprise to every department, employee and customer – heralds a turning point in which AI can move from being a technology used for specific use cases into one that defines the modern enterprise.
Chief information officers and technical leadership will need to take decisive action to embrace generative AI and seize the opportunities it offers. They will also have to make strategic decisions regarding data infrastructure, model ownership and workforce structure as well as AI governance, which will have a long-term impact on organizational success.
The report examines the current thinking of Chief Information Officers at some of the largest and most renowned companies in the world, as well experts from the private, public and academic sectors. The report presents their views on AI in the context of our global survey conducted with 600 senior data and tech executives.
The following are key findings:
* Unstructured data that was buried and unreadable is now readable, unlocking value for business. Prior AI initiatives were limited to use cases with abundant and ready structured data. The complexity of collecting and annotating heterogeneous datasets rendered wider AI initiatives infeasible. Generative AI's ability to uncover and use previously hidden data will drive extraordinary advances throughout the organization.
* To support the generative AI age, data infrastructures must be flexible, scalable and efficient. Chief information officers and technical leaders are adopting next-generation infrastructures to power these new initiatives. Data lakehouses are a more advanced approach that can improve security and allow for low-cost data storage and high-performance queries.
* Some organizations are leveraging open-source technologies to build their LLMs. They do this to protect their data and IP and to capitalize on it. CIOs understand the risks and limitations of using third-party services. This includes the risk of releasing sensitive information and the reliance on platforms that they don't control. Also, they see the potential of developing LLMs that are customized and maximizing value with smaller models. Successful organizations will find the right balance between risk, competitive advantage and governance.
* While dystopian predictions are not unfounded, automation anxiety is not something to ignore. The report interviewed CIOs and professors who do not anticipate a large-scale threat from automation. They believe that the workforce will be freed from tedious tasks and able to concentrate on more valuable areas such as insight, strategy, or business value.
AI will only progress if there is a unified and consistent governance. The risks of generative AI are commercial and societal, such as copyright infringements, toxic content, unreliable results or those that cannot be explained, and the protection of commercially sensitive IP. To be able to innovate without breaking anything or getting ahead of the regulatory changes, CIOs who are diligent must tackle the unique governance challenges that generative AI presents, by investing in technology and processes.
Download full report
The content of this article was created by Insights – the custom content division of MIT Technology Review. This content was not created by MIT Technology Review.
————————————————————————————————————————————————————————————
By: MIT Technology Review Insights
Title: The great acceleration: CIO perspectives on generative AI
Sourced From: www.technologyreview.com/2023/07/18/1076423/the-great-acceleration-cio-perspectives-on-generative-ai/
Published Date: Tue, 18 Jul 2023 13:00:00 +0000
Leave a Reply