Generative AI, the Great Productivity Booster?
After weeks of speculation, Microsoft announced last week that it is extending its partnership with OpenAI, the company behind ChatGPT, DALL-E 2 and GPT-3.
While all eyes have been on ChatGPT, the move marks a growing interest in generative AI in general. So how could such technology be used to support the digital workplace?
What Generative AI Offers
Matt White, CEO of Berkeley, Calif.-based Berkeley Synthetic, a generative AI research group, explained that generative AI is the ability for machines (more aptly deep-learning algorithms) to produce new data based on what they have learned from sampled data.
The generative model, he said, is trained on existing data. Take DALL-E 2 as an example. It is trained on hundreds of millions of image-text pairs, White said. This training allows the model to construct its knowledge of the concepts in the images (these are called parameters or weights).
This contrasts with discriminative AI, which is the ability to classify data and predict possible data points, an example of which is a fraud detection system used by credit card providers.
Generative AI is not just theoretical data science. “There will be far-reaching impacts to all industry verticals," White said, noting the technology has already been in use for some time, for things like language translation or text to speech.
While what has captured public attention recently are text-to-image applications (DALL-E 2, MidJourney, Stable Diffusion) and generative text or large language models like ChatGPT and GPT-3, White sees great potential for other emerging areas like text-to-video, text-to-3D and a myriad of other uses.
He offers three examples of generative AI's applications:
Healthcare: Generative AI could produce realistic scans of both cancer and non-cancer imagery that improve the accuracy of a deep learning (AI) model in detecting cancer in patients. This is called synthetic data generation, and it has life-changing potential.
Hyper-personalization: Marketing hyper-personalization is becoming more likely with generative AI. While targeted marking is already here, generative AI has the ability to enable marketers to produce unique or one-of-a-kind ads tailored to an individual’s preferences.
Call centers: Call centers may be necessary, but they're outdated, White said. Instead, generative AI can interpret audio, generate realistic voices, engage in realistic dialog and reference other non-generative AI systems to accomplish the most commonly encountered call center tasks — an area where interactive voice response (IVR) systems currently fail.
“What generative AI does is help enable the creator economy," White said. "One person will be able to do more than 50 [people] could do today — in a fraction of the time and cost."
There's no doubt, he said, that jobs and the workplace as we know it will be impacted, especially knowledge and entry-level jobs.
Related Article: What ChatGPT in Microsoft 365 Could Spell for the Workplace
How Does Generative AI Differ From Other AI?
The capabilities of generative AI are unmatched in the artificial intelligence sector, but does this mean it will replace other forms of AI?
PrasannaKumar Arikala, CTO of Orlando, Fla.-based Kore.ai, doesn't believe so. Conversational AI (CAI), for example, is integral to the development, deployment and success of virtual assistants in IT and HR settings, he said, and he doesn't see that going anywhere anytime soon.
Virtual assistants allow users to interact with technology using natural language conversations. They can be installed across the enterprise in various industries, including retail, banking, healthcare and HR environments, to improve user interactions, experience and outcomes. Generative AI can't do what CAI does.
The main distinction between CAI and generative AI is the element of human interaction. While CAI requires a person to draft and input pre-written responses to help guide the conversation flow, generative AI is designed to create new content based on the data it has been previously trained on with no direct human input.
This means that developers do not have direct control over what a model like ChatGPT says to its users, and generative AI on its own is not equipped to solve enterprise-specific problems like FAQs or queries unless specifically trained on that information.
“While ChatGPT can be used to generate responses that might be used by a chatbot or virtual assistant, it does not have enough features to be able to replace a conversational AI platform,” Arikala said.
Related Article: The Enterprise Chatbot: Your Future Coworker?
What Does Generative AI Mean for the Digital Workplace?
Using Arikala's comparison, we can say that in the digital workplace, generative AI models elevate the way virtual assistants are built and managed, thus eliminating the need for manually curating significant amounts of training data.
“Integration with these models can answer questions accurately from unstructured content sources such as FAQs, web pages and documents," Arikala said.
But the real magic of AI happens with the integration of generative AI tools into CAI-empowered virtual assistants, which Arikala said can greatly improve customer-facing interactions and aid workers.
Learning Opportunities
The bot development process is heavily simplified through the integration of generative AI. This includes the ability to identify new intents, design conversational flow use cases that cover a wide range of variations and generate training utterances and test cases for use cases.
Rosaria Silipo, head of data science evangelism at software developer KNIME, said the power of generative AI for the workplace includes all those algorithms that generate new content, like texts or images, from scratch. GANs (Generative Adversarial Networks), she said, are the best-known neural networks used for this task (more on that below).
In other words, for digital workplace workers, generative AI produces content that does not exist. It's easy to imagine what can be achieved with those capabilities, from generating stories, songs, even movie plots, to creating new models, prototypes, mockups, even images of people who do not exist or of inexistent landscapes.
Generative AI can also be used to enhance creativity. Workers can use it to brainstorm and generate ideas. But although it may give a false sense of security that it is completely original output, some experts are cautioning that this use of AI may cause potential copyright issues. There is also risk for fake content and fake identities, which are important considerations to keep in mind for companies seeking to tap into the potential of generative AI.
Related Article: How to Spot Deep Fake Remote Workers
Increasing Productivity? Maybe
Given the novelty of generative AI, the productivity argument may sound a bit hypothetical. After all, there isn't a long history of testing on which to base any kind of case.
But Sugandha Sahay, technical program manager at AWS, argues that adding generative AI tools to the digital workplace can significantly increase workers' productivity. And research by Sequoia Capital VC has found that generative AI does have the potential to make digital workers at least 10% more efficient or creative.
Sahay offered a number of examples of how generative AI could support increased productivity in the workplace:
- Delegating tasks like creating meeting minutes.
- Developing user-friendly product documentation like user or reference guides.
- Conceptualizing media-rich user experience walkthroughs.
- Personalizing marketing campaigns and advertisements.
- Generating leadership dashboards and stakeholder reports.
Sahay said generative AI has the potential to not only enhance productivity but also efficiency, from supporting content creation to optimizing processes. For instance, a generative AI model can be used to optimize scheduling, routing or supply chain management, which could lead to increased efficiency, reduced costs and improved overall performance for the business, she said.
“AI has the potential to greatly benefit the digital workplace by enhancing productivity, efficiency and innovation," Sahay said.
Related Article: ChatGPT Shows AI Is a Better Co-Pilot Than Solo Driver at Work
The Potential Behind Generative AI Models
Generative AI is not a single AI model. Rather, as Sharda Kumari, a systems architect with AirbnB notes, there are different varieties of generative models. Two major examples are GANs and variational autoencoders (VAE).
GANs consist of two neural networks: a generator network and a discriminator network. The generator network is trained to generate new data samples that are similar to the ones in the training set, Kumari said, while the discriminator network is trained to distinguish between real and generated data samples. During training, the generator and discriminator networks are optimized simultaneously, with the generator trying to generate realistic data samples that can fool the discriminator, and the discriminator trying to correctly identify whether a data sample is real or generated.
VAEs, on the other hand, are often used for image and audio generation tasks. They consist of an encoder network and a decoder network. The encoder network is trained to map the input data to a lower-dimensional latent space, while the decoder network is trained to map the latent space back to the original data space.
These are happening now, already. But Kumari said there are so many more possibilities on the horizon.
“As the field of generative AI continues to evolve, it is likely that even more powerful and versatile models will be developed, which will open up new possibilities for the digital workplace," she said.
About the Author
David is a full-time journalist based in Ireland. A partisan of ‘green’ living and conservation, he is particularly interested in information management and how enterprise content management, analytics, big data and cloud computing impact on it.
Featured Research
Research Report
Read Now
eBook
Read Now
Guide
Read Now
Guide
Read Now
Research Report
Read Now
Buyer's Guide
Read Now
Infographic
Read Now
White Paper
Read Now
Related Stories
Smarter Experiences Win — Learn from Today's Leaders
