The Evolution and Impact of Advanced AI: From GPT-4 to Enterprise Applications

Council Post: GPT-4 And Beyond: The Power And Potential Of Advanced Generative AI

Can artificial intelligence (AI) and natural language processing (NLP) reach the next level of sophistication and nuance? At our company, we've been exploring the potential of advanced generative AI (GenAI), such as GPT-4, to achieve just that. This latest iteration from OpenAI has already surpassed its predecessors, setting a new standard in language modeling. In this article, I want to share our experiences using this sophisticated AI, our challenges and how we've overcome them, offering insights that could help other businesses enhance their strategies. The current state of advanced generative AI demonstrates impressive capabilities in understanding and maintaining context, generating high-quality text and providing personalized experiences. These developments are crucial for widespread adoption, reflecting significant strides in AI's ability to sound more human and handle complex queries with nuanced responses. GPT-4, for example, boasts significant improvements, including a beefed-up neural network for greater accuracy and nuance in text, and enhanced training on a substantial dataset to expand its language and contextual awareness. The standout feature, however, is its context-keeping ability. GPT-4 can reason through longer passages of text, making conversations sound much more human. It handles complex or ambiguous queries robustly, tailoring responses to user requests. These advancements have been pivotal in our use of GPT-4 for various applications within our company. In the business world, generative AI like GPT-4 can be leveraged for a multitude of tasks. In our software and video game development company, Gameverse, we have harnessed GenAI technologies in several innovative ways. For instance, we use GPT-4 to help fine-tune dynamic, engaging dialogues for our video games, providing players with a more immersive experience. Additionally, GPT-4 helps in automating code generation and debugging, significantly speeding up the development process. These specific applications showcase how GenAI can push the boundaries of creativity and efficiency in the gaming industry. Implementing GPT-4 was not without its challenges. One significant hurdle was ensuring data privacy and security. We adopted stringent data protection policies and employed a dedicated team to monitor and mitigate risks associated with bias and hateful content. Initially, our primary challenge was integrating these stringent measures without disrupting our workflow. The team, consisting of data scientists, cybersecurity experts and AI ethicists, was crucial in developing robust protocols that balanced security and efficiency. Based on our experience, we recommend that other companies tailor their approach to their specific needs and resources, ensuring they allocate sufficient expertise to handle these critical aspects. Integrating GPT-4 into our existing systems also required substantial training and fine-tuning. By focusing on continuous learning and adaptation, we were able to overcome these challenges, ensuring the technology could meet our specific needs. This involved developing tailored training programs for our team, setting clear adaptation goals and constantly monitoring performance to make necessary adjustments. Our lead data scientist spearheaded this initiative, emphasizing collaboration and open communication to get everyone on board with the changes. This approach was essential for maintaining high performance and morale. Finally, deploying AI technologies like GPT-4 responsibly is paramount. Our company has implemented measures to ensure ethical use, including transparency in AI operations and adherence to privacy regulations. OpenAI's commitment to safety, including rigorous moderation to detect and mitigate biases, aligns with our approach. However, broader societal questions around AI ethics necessitate ongoing debate and regulation to ensure democratized and transparent use. The future for GPT-4 and other AI technologies appears promising. Machine learning and neural network advances will likely yield even more sophisticated models. Future iterations may introduce real-time interaction improvements; multi-modal models integrating text, image and audio; and more personalized AI experiences. As AI continues to evolve, it will become increasingly integral to our daily lives, transforming various aspects of human activity. Our experiences with GPT-4 highlight its potential to revolutionize business processes and beyond. This technology represents a milestone in AI evolution, offering practical applications that enhance efficiency and innovation. As we navigate this rapidly advancing landscape, overcoming challenges such as ethical considerations remains critical to ensuring AI technologies serve as a force for good. By sharing our journey, we hope to provide valuable insights for other organizations looking to leverage the power of generative AI. Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?

Council Post: More Than Just GenAI: How Harnessing The Full Spectrum Of AI Can Revolutionize Customer Experiences

Despite the global buzz around large language models (LLMs), misconceptions about AI persist. AI is not one single thing. There are three main types: perceptive AI, analytical AI and generative AI (GenAI). Each can help companies solve different customer experience (CX) challenges in meaningful ways. While LLMs like ChatGPT have brought GenAI into the mainstream, business leaders should understand the bigger picture. This is not to say GenAI isn't important or remarkably versatile. Rather, it is just one tool that can help address the daunting challenge of improving customer experiences. Over the last 20 years, automation has improved CX in many ways. It has given consumers new self-help options -- and enabled them to solve their problems more quickly. But they're not yet perfect. Everyone knows the frustrations of time-wasting friction in many self-service applications. Now, advances in AI promise to reduce these lingering CX challenges. To find out how it first helps to understand the three types of AI and how you can apply them. AI is already transforming CX across all industries. In the telco sector, for example, mobile operators use AI to reduce pain points and improve satisfaction in device-related journeys. A device-related journey starts when a customer upgrades to a new device -- a smartphone, for instance -- either through a trade-in or a direct purchase. It continues with the onboarding process, which often involves transferring content to the new device and wiping the old one clean. The journey then extends through device care, like troubleshooting device-related issues or fixing a broken screen, and concludes with another trade-in upgrade, creating an ongoing cycle. For mobile operators, these journeys are critical. They often start with a positive experience, boosted by promotions and the excitement of a new device. However, any part of the journey has the potential to turn negative. Regrettably, they often do -- with very negative consequences. Today, mobile devices are deeply woven into people's interactions, social connections and routines. Even a minor malfunction can cause immense frustration, especially when trying to resolve the issue turns into an even bigger headache. There is tremendous pressure on mobile operators to resolve CX problems to avoid customer churn and a drop in net promoter scores (NPS). Using the right types of AI for the right use cases can effectively address customer pain points. It can turn a subjective, manual and inconclusive experience into an objective, automated and conclusive experience that strengthens a customer's satisfaction and loyalty. Best practice: Combine perceptive and analytical AI to bring expert-level assessment capabilities directly into customers' hands. This removes friction and uncertainty, leading to improved trust and satisfaction. Example: Customers rarely get a reliable estimate when trading in their mobile device. According to MCE's survey of 23,000 mobile customers, half of the respondents said they received a different trade-in credit from the original quote. In one case, simply implementing a price guarantee reduced one-third of complaints and calls to customer care. Here's where AI can bring warehouse "intelligence" directly to consumers' fingertips. By combining perceptive AI with analytical AI, mobile operators can objectively assess the key value drivers of a device -- the camera and the screen -- to offer customers a guaranteed trade-in value in minutes. * Camera: Perceptive AI assesses the camera's working condition by using perceptive AI's object detection capabilities. * Cosmetic condition: The consumer holds the device up to a mirror, and perceptive AI captures the screen's quality by identifying scratches, cracks and chips. * Overall grade: Analytical AI translates these data points into an objective measurement of the overall device condition. And, using additional process device data, calculates a grade and a final trade-in value. Best practice: Develop AI-powered tools that quickly diagnose issues and provide personalized solutions that minimize wait times and boost customer satisfaction. Example: It isn't surprising that any device care event drops an operator's NPS by 19, according to our research. While most customers would prefer to diagnose and fix issues on their own, less than 30% have used a self-service tool. Even with customer support in retail or call centers, 10% of customers still cannot resolve their problems, leading to a severe drop of 36 points in NPS. One-third don't get personalized alternative solutions, even though customers appreciate receiving them during a device care journey. The powerful capabilities of analytical AI can allow customers to access device diagnostic test suites through an app. Analytical AI can also ingest additional data sources regarding the device and the customer (device make, model, CRM data, etc.), process the outcomes of the diagnostic tests, and make suggestions. Our experience has found that about 40% of issues can be resolved in the app alone through tutorials, automatic settings changes or "next-best actions" for alternative solutions. For example, after asking the customer to complete a series of tests, AI might recommend a trade-in instead of a repair, helping them avoid a long and complicated repair process, thus improving their satisfaction. Best practice: Incorporate generative AI to create a conversational interface that can handle complex queries and offer tailored solutions. Example: Generative AI can act as the ultimate customer service agent, empowering customers to solve most device-related issues through a conversation augmented with real-time device diagnostics and insights driven by analytical AI. GenAI's personalized approach helps users solve even more complex issues while enabling seamless integration of commercial offers that often yield higher conversion rates. CX is more important than ever in today's hypercompetitive business environment. In many sectors, customer satisfaction is the difference between success and failure. The three flavors of AI present companies in all industries with an unprecedented opportunity. For the first time, they can combine AI's ability to perceive, analyze and generate to solve customer problems quickly and with minimal friction. Our experience in the telco space demonstrates the power of AI in CX. By addressing specific pain points around device journeys, mobile operators can reduce the uncertainty that often leads to distrust and replace it with higher satisfaction and loyalty. Forbes Communications Council is an invitation-only community for executives in successful public relations, media strategy, creative and advertising agencies. Do I qualify?

Enterprises embrace generative AI, but challenges remain

We want to hear from you! Take our quick AI survey and share your insights on the current state of AI, how you're implementing it, and what you expect to see in the future. Learn More Less than two years after the release of ChatGPT, enterprises are showing keen interest in using generative AI in their operations and products. A new survey conducted by Dataiku and Cognizant, polling 200 senior analytics and IT leaders at enterprise companies globally, reveals that most organizations are spending hefty amounts to either explore generative AI use cases or have already implemented them in production. However, the path to full adoption and productivity is not without its hurdles, and these challenges provide opportunities for companies that provide generative AI services. Significant investments in generative AI The survey results announced at VB Transform today highlight substantial financial commitments to generative AI initiatives. Nearly three-fourths (73%) of respondents plan to spend more than $500,000 on generative AI in the next 12 months, with almost half (46%) allocating more than $1 million. However, only one-third of the surveyed organizations have a specific budget dedicated to generative AI initiatives. More than half are funding their generative AI projects from other sources, including IT, data science or analytics budgets. It is not clear how pouring money into generative AI is affecting departments that could have otherwise benefitted from the budget, and the return on investment (ROI) for these expenditures remains unclear. But there's optimism that the added value will eventually justify the costs as there seems to be no slowing in the advances of large language models (LLMs) and other generative models. "As more LLM use cases and applications emerge across the enterprise, IT teams need a way to easily monitor both performance and cost to get the most out of their investments and identify problematic usage patterns before they have a huge impact on the bottom line," the study reads in part. A previous survey by Dataiku shows that enterprises are exploring all kinds of applications, ranging from enhancing customer experience to improving internal operations such as software development and data analytics. Persistent challenges in implementing generative AI Despite the enthusiasm around generative AI, integration is easier said than done. Most of the respondents in the survey reported having infrastructure barriers in using LLMs in the way that they would like. On top of that, they face other challenges, including regulatory compliance with regional legislation such as the EU AI Act and internal policy challenges. Operational costs of generative models also remain a barrier. Hosted LLM services such as Microsoft Azure ML, Amazon Bedrock and OpenAI API remain popular choices for exploring and producing generative AI within organizations. These services are easy to use and abstract away the technical difficulties of setting up GPU clusters and inference engines. However, their token-based pricing model also makes it difficult for CIOs to manage the costs of generative AI projects at scale. Alternatively, organizations can use self-hosted open-source LLMs, which can meet the needs of enterprise applications and significantly cut inference costs. But they require upfront spending and in-house technical talent that many organizations don't have. Tech stack complications further hinder generative AI adoption. A staggering 60% of respondents reported using more than five tools or pieces of software for each step in the analytics and AI lifecycle, from data ingestion to MLOps and LLMOps. Data challenges The advent of generative AI hasn't eliminated pre-existing data challenges in machine learning projects. In fact, data quality and usability remain the biggest data infrastructure challenges faced by IT leaders, with 45% citing it as their main concern. This is followed by data access issues, mentioned by 27% of respondents. Most organizations are sitting on a rich pile of data, but their data infrastructure was created before the age of generative AI and without taking machine learning into account. The data often exists in different silos and is stored in different formats that are incompatible with each other. It needs to be preprocessed, cleaned, anonymized, and consolidated before it can be used for machine learning purposes. Data engineering and data ownership management continue to remain important challenges for most machine learning and AI projects. "Even with all of the tools organizations have at their disposal today, people still have not mastered data quality (as well as usability, meaning is it fit for purpose and does it suit the users' needs?)," the study reads. "It's almost ironic that the biggest modern data stack challenge is ... actually not very modern at all." Opportunities amid challenges "The reality is that generative AI will continue to shift and evolve, with different technologies and providers coming and going. How can IT leaders get in the game while also staying agile to what's next?" said Conor Jensen, Field CDO of Dataiku. "All eyes are on whether this challenge -- in addition to spiraling costs and other risks -- will eclipse the value production of generative AI." As generative AI continues to transition from exploratory projects to the technology underlying scalable operations, companies that provide generative AI services can support enterprises and developers with better tools and platforms. As the technology matures, there will be plenty of opportunities to simplify the tech and data stacks for generative AI projects to reduce the complexity of integration and help developers focus on solving problems and delivering value. Enterprises can also prepare themselves for the wave of generative AI technologies even if they are not exploring the technology yet. By running small pilot projects and experimenting with new technologies, organizations can find pain points in their data infrastructure and policies and start preparing for the future. At the same time, they can start building in-house skills to make sure they have more options and be better positioned to harness the technology's full potential and drive innovation in their respective industries.

Beyond the gen AI hype: Google Cloud shares key learnings

Is bigger always better when it comes to large language models (LLMs)? "Well, the answer is quite simply yes and no," Yasmeen Ahmad, managing director of strategy and outbound product management for data, analytics and AI at Google Cloud, said onstage at VB Transform this week. LLMs do get better with size -- but not indefinitely, she pointed out. Huge models with a large number of parameters can be outperformed by smaller models trained on domain and context-specific information. "That indicates that data is at the cornerstone, with domain-specific industry information giving models power," said Ahmad. This allows enterprises to be more creative, efficient and inclusive, she said. They can tap into data that they've never been able to access before, "truly reach" all corners of their organization and enable their people to engage in all new ways. "Gen AI is pushing the boundaries of what we could even dream machines could create, or humans could imagine," said Ahmad. "It truly is blurring the lines of technology and magic -- perhaps even redefining what magic means." Enterprises need a new AI foundation Successfully training models on a specific enterprise domain comes down to two specific techniques: fine-tuning and retrieval augmented generation (RAG), said Ahmad. Fine-tuning teaches LLMs "the language of your business," while RAG allows the model to have a real-time connection to data, whether in documents, databases or elsewhere. "That means in real-time, it can provide accurate answers which are really important for financial analytics, risk analytics and other applications," said Ahmad. Similarly, the true power of LLMs is in their multimodal capabilities, or their ability to operate on video, image, text documents and all other types of data. This is critical, she noted, as typically 80 to 90% of data in an enterprise is multimodal. "It's not structured, it's documents, it's images, it's videos," said Ahmad. "So having a LLM to be able to tap into that data is super valuable." In fact, Google did a study that showed a 20 to 30% improvement in customer experience when multimodal data was used. Enterprises had enhanced ability to hear and understand customer sentiment and the model was able to bring together data on product performance and market trends. "To put it simply, it's not about simple pattern recognition anymore," said Ahmad. "LLMs can truly understand the complexity of our organizations by having access to all data." Traditional organizations struggle with traditional data foundations that were never built to handle multimodal -- but the future of AI and business data demands a new kind of AI foundation, she pointed out. AI that is conversational, a 'personal data sidekick' The ability to engage in question-answer interactions is another critical component of successful LLMs, Ahmad emphasized. But, while it's "super alluring to be able to chat with your business data, it's not so easy," she noted. Imagine asking a colleague the forecasted sales for the next quarter for new products. If you don't give them context, or if they don't understand the fiscal quarters or even the new products themselves, they are going to give you a "vague and unhelpful" answer, said Ahmad. The same is true for LLMs -- they must be given semantic context and metadata so they can provide specific and accurate answers. Similarly, it's important that models are conversational. "As humans, when we do analysis, or we ask questions, we typically go back and forward in a dialog, and we call on and provide additional context until we get to an answer," said Ahmad. It's exactly the same for LLMs: They need to be able to have a coherent conversation. As such, the industry is moving away from isolated, single-shot, one question interactions to "the next generation of conversational AI." This is more than a chatbot: "Think of it more like a personal data sidekick," she said. It is a "tireless worker" that interacts and is able to ask questions and engage in a chain of thought. It also provides thorough query transparency, so human users know where the results came from and can trust them. "We're seeing a quantum leap, agentic AI that can actually make decisions, take action and work towards a goal," said Ahmad, noting that scientists are teaching these models to become "seriously clever." LLMs are beginning to mimic human brains -- notably in the way they can break things into sub tasks -- and they have the ability to be "strategic thinkers," understand cause and effect and learn honesty. All of this is being done quicker and quicker, with real-time capabilities improving all the time, said Ahmad. "The future is here and the future is spawning new breeds of business," she said. "We are at the beginning of what this technology can enable."

Determining the right LLM for your organization

Today's business leaders recognize that some application of generative AI has great potential to help their business perform better, although they may still be exploring exactly how and what the ROI may ultimately be. Indeed, as companies turn their gen-AI prototypes into scaled solutions, they must take into account such factors as the technology's cost, accuracy, and latency to determine its long-term value. The growing landscape of large language models (LLMs), combined with the fear of making the wrong decision, leaves some businesses in a quandary. LLMs come in all shapes and sizes and can serve different purposes, and the truth is, no single LLM will solve every problem. So, how can a business determine which one is the right one? Here, we discuss how to make the best selection so your business can use generative AI with confidence. Some businesses are conservative in adopting an LLM, launching pilot projects, and waiting for the next generation to see how that might change their application of generative AI. Their reluctance to commit may be warranted, as diving in too early and failing to test it correctly could mean big losses. But generative AI is a rapidly evolving technology, with new foundational models introduced seemingly weekly, so being too conservative and continuing to wait for the technology to evolve may mean you never actually move forward. With that said, there are three levels of sophistication companies may consider when it comes to generative AI. The first is a simple wrapper application around GPT, designed to interact with OpenAI's language models and provide an interface to guide text completions and conversation-based interactions. The next level of sophistication is using an LLM with retrieval-augmented generation (RAG). RAG allows businesses to enhance their LLM output with proprietary and/or private data. GPT-4, for example, is a powerful LLM that can understand nuanced language and even reasoning. However, it hasn't been trained on the data for any specific company and can lead to potential inaccuracies, inconsistencies, or irrelevancies (hallucinations). Companies can get around hallucinations by using implementations like RAG, which allows them to merge insights from a base-model LLM with some of the data unique to their business. (It should be noted that alternative large-context models like Claude 3 may actually render RAG obsolete. And, while many are still in their infancy, we all know how fast technology moves, so obsolescence may come sooner than later.) In the third level of generative AI sophistication, a company runs its own models. For example, a company may take an open-source model, fine-tune it with proprietary data, and run the model on its own IT infrastructure in place of any third-party offerings like OpenAI. It should be noted that this third-level LLM requires the oversight of engineers trained in machine learning. Given the options here and the differences in cost and capability, companies must determine exactly what they plan to accomplish with their LLM. For example, if you're an ecommerce company, human support is trained to intervene when a customer is at risk of abandoning their cart and help them decide to complete their purchase. A chat interface will allow for the same result at one-tenth the cost. In this case, it may be worth it for the ecommerce company to invest in running its own LLM with engineers to control it. But bigger isn't always cost-effective -- or even needed. If you're a banking application, you can't afford to make transaction errors. For this reason, you'll want tighter control. Developing your own model or using an open-source model, fine-tuning it, applying heavily engineered input and output filters, and hosting it yourself gives you all the control you need. And for those companies that simply want to optimize the quality of their customers' experience, a well-performing LLM from a third-party vendor would work well. Regardless of the chosen LLM, understanding how the model performs is key. As tech stacks become increasingly complex, homing in on performance issues that may pop up in an LLM can prove challenging. Additionally, due to the uniqueness of the tech stack and the very different LLM interactions, there are entirely new metrics that must be tracked, such as time-to-token, hallucinations, bias, and drift. That's where observability comes into play, providing end-to-end visibility across the stack to ensure uptime, reliability, and operational efficiency. In short, adding an LLM without visibility could greatly impact how a company measures the technology's ROI. The generative AI journey is exciting and fast-paced -- if not a bit daunting. Understanding your business's needs and matching those to the right LLM will not only ensure short-term benefits but also lay the foundation for ideal future business outcomes. We feature the best AI tools.

Making generative AI human-centric

We want to hear from you! Take our quick AI survey and share your insights on the current state of AI, how you're implementing it, and what you expect to see in the future. Learn More Building tools with generative AI has never been easier, but developers must also remember to experiment and focus on human end-users. Jessica Gilmartin, chief revenue officer at Calendly, said during the Women in AI Breakfast at VB Transform today that while so much of generative AI is exciting, the technology shouldn't be the goal at the end of the day. "AI is great, but you have to build for humans and work with them. Technology enables change; it cannot be the end result," Gilmartin said. She added that thinking of AI tools as human-first tools brings people of different backgrounds into the development process. Gilmartin said that at Calendly, the company encourages teams to include people with different expertise to collaborate. Other speakers echoed the idea that AI still needs a human in the loop, especially if the end product is meant to be used by people anyway. However, it isn't always easy to convince people that they can play a role in bringing generative AI applications to life. Encouraging experimentation "Lots of people don't have a generative AI background because it's new. I believe that you can use your background if it's in education, English, legal, you can play a role in AI development, you just need to experiment with it and actually use the technology," said Aparna Sinha, head of AI Product at Capital One. Acknowledging the speed of development, LinkedIn's head of Data and AI, Ya Xu, said, "There's no better time to jump into AI" than now, with several tools, podcasts and videos pointing to the best research papers to get up to speed now available. Kari Briski, vice president of AI models, software, and services at Nvidia, pointed out it's essential to carve out the time to learn tools and play around with AI. What is clear, the speakers said, is that people have to become comfortable with AI tools and in the process of building applications with the technology, it will only become better if more people of different backgrounds participate in developing it.