The financial industry will be the first to learn about GPT-3 • The Register
It’s hard to ignore the buzz around massive language models like GPT-3. It’s not your typical natural language processing (NLP) engines that power enterprise chatbots or call centers. This is a dramatic step forward, making traditional NLP output look like a simple rollback of trained responses.
GPT-3 can generate its own articles. It can provide nuanced summaries of millions of words of text. And since the code is essentially a language, there are now software elements created by GPT-3, including fully developed applications. It’s the closest thing we have yet that resembles the “omniscient supercomputer” trope of mid-20th century science fiction: the thing that you could actually ask a verbal question and receive a nuanced answer. .
So this is the third time you’ll see the word nuanced, because that’s where the power lies for companies that can actually access GPT-3 and massive training computing power. It is, after all, not unreasonable to see billion-dollar model-training efforts on the horizon.
Besides Meta and Google and the social giants, the GPT-3 production will likely be the first where users can afford this kind of return on investment (RoI). Healthcare might seem like a natural first option, but the regulatory environment isn’t too supportive of black box technologies. The financial services segment is where we’re likely to see it first, at least in consumer-facing businesses.
The use cases for GPT-3 in financial services are vast and already included in specific machine learning packages. For example, sentiment analysis (using social media and articles to capture market temperature), entity recognition (classifying documents), and translation are all widely available and used.
Where GPT-3 will likely come into play for banks is in language generation – the ability to handle complaints and fill in information in forms, for example. It might be a small, consumer-focused start, but with enough training data, GPT-3 could begin to play an active role in risk management and investment decisions.
It is difficult to master the current return on investment of this technology in the banking sector. These ML elements exist, but as data volumes grow, the need for massive industry models and even bank-specific trained models becomes clearer. A big problem for financial institutions able to access the model (OpenAI is less closed these days but GPT-3 is limited in terms of pre-training, adjustment of tasks downstream, no longer any corpus specific to the industry) is finding the people to make everything work.
Yes, the irony is that something designed to replace people is slow to catch on due to a lack of staff. That’s why, like so much else in enterprise IT, the use cases for managed services abound. Finding and retaining talent is harder than ever, especially for specific AI/ML financial services programs. Plus, it’s no small feat to have a bank-specific corpus pre-trained and ready to go with the right sized system and software.
GPT-3 models have been available to a select few since mid-2020 through OpenAI and the sole licensee is Microsoft. It didn’t start out as an AI framework that the general public can just grab and work with, which is why early system makers to have it tuned with their machines and software stacks have a head start. . This reduces the hiring problem in AI and puts all the tools at your fingertips. However, few companies do.
AI systems maker SambaNova was among the first to deliver GPT-3 as a pre-packaged service earlier this year so companies can start digging into building and using large language models. . Among their target markets are financial services, a segment that can leverage the capabilities of GPT-3 by blending data to power hedge funds and risk decision-making, for example. The sole public user of their GPT Banking platform, OTP Bank in Hungary, continues to launch tires with a prototype rollout.
Large banks exploring GPT-3 on a large scale may be cautious in terms of investment. SambaNova’s Product SVP said The register that deployments can be on-premises hybrid cloud (the right-sized hardware behind the firewall, owned and managed by SambaNova and accessed through their APIs) or they can use it as a CSP.
Either way, it’s a subscription service, which at this point in calculating the ROI of GPT-3 means that only the financial industry can determine if the hype is right placed for the large-scale GPT-3 in the enterprise. ®