2024 State of AI in the Speech Technology Industry: AI’s Impact on Natural Language Processing
When it comes to speech technologies, natural language processing (NLP)—the capacity for computers to understand human language as it is written and spoken—has been a challenge. But thanks to recent breakthroughs, AI has dramatically enhanced NLP.
The progression from early language models to advanced systems has led to a significant leap in machine comprehension and text generation, facilitating nuanced and context-aware interactions. Tools that leverage NLP are increasingly complex and user-friendly enough to address our repetitive, daily tasks and solve more complex issues. This has had a profound impact on various industries, improving customer service through intuitive chatbots, streamlining the analysis of healthcare records and communication with patients, and providing more personalized experiences for retail customers.
Notably, generative AI—a form of AI that can create different types of content, such as text, audio, images, and synthetic data—has emerged as a pivotal factor in elevating NLP by enabling the comprehension and generation of language and text that closely resembles human expression.
Phil Portman, founder and CEO of Textdrip, sees AI playing an essential role in enhancing NLP capabilities, enabling more accurate language understanding and generation.
“From sentiment analysis to language translation, AI has significantly improved the efficiency and effectiveness of NLP applications,” he says.
OpenAI’s ChatGPT, launched in late 2022 and now in its fourth iteration, made the general public aware of what large language models (LLMs) and generative AI can do, demonstrating that AI could respond in a humanlike way to questions or requests.
“Now, other companies are working on improving their LLMs, including Google and Amazon, and more businesses are adopting AI technologies to improve their online chatbots, analyze their data, and help their engineers develop software faster,” says Maria Kyrarini, assistant professor of electrical and computer engineering at Santa Clara University in California.
Intelligent voice-controlled assistants like Amazon’s Alexa, Apple’s Siri, and Google Assistant have long used NLP to answer simple, everyday consumer questions, and contact center chatbots today better understand natural language communicated through customer interactions. But AI and NLP have gone far beyond chatbots.
“In the contact center today, for instance, NLP and AI models can deliver real-time guidance to agents while interacting with customers,” notes Josh Feast, cofounder and CEO of Cogito. “The technology can identify and predict the next steps within the complex, human-led interactions.”
But the road to this point has hardly been smooth. From early NLP research in the 1930s, the AI space has witnessed a handful of significant crashes and downturns as overly ambitious claims collided with processing power constraints.
“Anyone tuned in to the history of NLP is painfully aware of the roller coaster of AI research: spiraling hype cycles followed by inevitable periods of AI winter where funding has collapsed and interest in AI has dried up for years at a time,” explains Amanda Robinson, director of conversational AI and customer experience design at TTEC Digital. “Currently, we’re in the midst of an AI boom which arguably started around 2011 and accelerated dramatically in 2022 with the public launch of generative AI tools powered by LLMs.”
Before that crucial year, software engineers frequently encountered challenges in performing tasks that are now considered simple, like using AI to summarize documents or compile slides from information. Today, not only can software seamlessly execute these tasks through application programming interfaces, but users also can interact with AI in remarkably nuanced ways.
“Someone can ask generative AI services to be more articulate, pithy, or concise, and the algorithms will produce an impressive result the majority of the time,” says Tony Lambert, chief technology officer of TechSmith. “The next step is to improve upon this groundwork and operationalize AI in business and consumer situations where it’s no longer a novelty and instead seamlessly embedded in software.”
Alex Lartey, chief business development officer at Tovie AI, is particularly bullish on near-future opportunities.
“AI has catalyzed a major shift in NLP from academic conjecture to functional, real-world tools, ushering in AI models that perceive context, grasp nuances, and parse deeper meanings in human language with impressive accuracy,” he says. “AI in NLP now facilitates near-human understanding in chatbots, can outperform traditional analyses in data interpretation, and bridges communication gaps, enhancing efficiency and spawning new opportunities for innovation across multiple sectors.”
In 2024, experts anticipate that AI-driven NLP will elevate the management of intricate dialogues, provide enhanced language flexibility, and advance sentiment analysis. Businesses will also incorporate more sophisticated AI-driven personalization, and new models will continue to narrow the divide between human and machine communication, experts agree.
But before we look too far ahead, let’s recap some of the biggest AI-NLP headlines of last year, which saw advancements in grasping context, the development of advanced chatbots, the merging of AI with voice technology, and prominent models and contributions by industry giants like OpenAI, Google, and Microsoft.
“In 2023, the big news was OpenAI debuting ChatGPT-4 and Meta debuting Llama,” Feast says. “Organizations increasingly realized that AI-powered NLP can revolutionize business and, specifically, contact center operations by automating routine tasks, increasing employee efficiency, and improving the customer experience.”
Another significant 2023 milestone was the development of smaller LLMs like Google’s Gemini, a multimodal AI model.
“Multimodal refers to models that leverage images, videos, audio, and words simultaneously to produce an output as opposed to primarily text-to-text AI,” Lambert says. “Now, users can ask AI to create a picture that looks a certain way and refine it incrementally or provide a picture of refrigerator items and ask it to create a recipe. This is something that didn’t exist to consumers even a year ago and is essential to the widespread adoption of AI.”
Lartey points to other significant developments over the past several months, including the following:
- Sentiment analysis, which now captures text emotions and opinions with high fidelity, reshaping social media scrutiny, feedback analysis, and market research by delving deep into collective sentiments.
- Better refined named entity recognition (NER), which excels at pinpointing and categorizing text data, transforming unstructured data into orderly, accessible information for retrieval and organization tasks.
- Improved language transformer models that are honing text generation, translation, and summarization capabilities, continuously expanding AI’s linguistic potential.
- The evolution of virtual assistants, which boast better context recognition and humanlike interactions, revolutionizing customer service, personal assistants, and the smart home industry.
Amitha Pulijala, vice president of product at Vonage, is encouraged that AI and NLP are becoming tightly woven across several industries, from retail to healthcare and beyond.
“We’re seeing chatbots conversing with uncanny nuance, algorithms that understand and generate language with newfound depth, and digital assistants that can personalize experiences with a little bit of AI magic,” she says.
A Look Ahead
Fortune Business Insights has projected the NLP market to grow from $24.1 billion in 2023 to $112.3 billion by 2030, expanding at a compound annual growth rate (CAGR) of 24.6 percent. MarketsandMarkets expects the speech and voice recognition market to reach $28.1 billion by 2027, growing at a CAGR of 24.4 percent.
“These statistics underscore the immense market potential and growing recognition of AI and NLP’s transformative impact on industries worldwide,” Portman says.
Dave Hoekstra, product evangelist at Calabrio, believes NLP powered by AI has reached its highest saturation point yet.
“However, adoption has been challenging because humans innately distrust computers to think the same way a human mind would, and businesses are reluctant to spend real capital on something that may cost them long term,” he says.
But for all the gains, several sizable hurdles must be cleared in the months and years ahead. Robinson foresees five major challenges:
- Unauthorized use of copyrighted materials or classified personal information in training LLMs.
- Bias, such as lack of representation, stereotypes, toxic content, discrimination, racism, sexism, and hate speech.
- Hallucinations. “All LLMs hallucinate to some degree via incorrect facts, loss of context, distortions of the source material, and even making things up. Before LLM-based chat solutions can be deployed for customer-facing interactions, we must be able to trust the answers,” Robinson cautions.
- Ongoing security risks and data privacy as hackers persistently discover new exploits through prompt engineering and other methods.
- Environmental impact. “The footprint of training and querying LLMs keeps expanding as the number of model parameters continues to grow,” Robinson says. “LLMs require a tremendous amount of computing power and electricity, with the resulting carbon dioxide emissions creating serious negative impacts on the environment.”
“Many businesses are also still figuring out how to effectively integrate or apply AI into their operations to solve specific business problems,” Pulijala says. “Balancing the impact of human-AI collaboration is also a challenge as companies work to ensure that AI is augmenting human capabilities vs. replacing them.”
Another barrier is the lack of clean, quality datasets to train NLP models. Feast says this can take a lot of time to resolve and not all companies can dedicate resources to this.
Still, Pulijala is convinced the market for AI and NLP tools will continue to grow as industries continue to find new ways to use these technologies.
“Take healthcare alone, where AI and NLP will continue to improve patient care, create more personalized treatment plans, analyze large sets of patient data, and produce more efficient communications with healthcare providers. And the customer service industry will see the growth of AI-powered chatbots and virtual assistants as they become even more human-like and able to handle more in-depth requests,” she says.
Maurice Kroon, CEO and founder of Vox AI, foresees a surge of startups and niche companies bringing fresh, innovative solutions to the table, and smarter, more nuanced AI that handles the subtleties of human language even better.
“We’re not just talking about understanding different languages but also the cultural and contextual layers within them,” he says.
Innovations visible in Portman’s crystal ball include improved model efficiency, reduced bias, and more seamless integration of AI with NLP applications. “Continued advancements in transformer architectures and unsupervised learning will shape the technological landscape, too,” Portman says.
Additionally, expect a new wave of AI assistants “that support software users at the moment they require it instead of having to go to an outside source,” Lambert predicts.
Generative AI will continue to drive a platform shift in content creation and knowledge generation, and continued LLM improvements will further accelerate and democratize the adoption of AI-driven NLP, Pulijala predicts.
“We will also see the embedding of NLP models into existing [software-as-a-service] platforms and day-to-day solutions. Rather than searching by terms, we will start using Google to search according to our specific questions,” Feast maintains.
The vendor landscape will also likely see increased competition, with both established players and startups contributing to a diverse ecosystem.
“Collaboration and partnerships will be key as vendors aim to offer comprehensive solutions that cater to evolving industry needs,” Portman portends.
Erik J. Martin is a Chicago area-based freelance writer and public relations expert whose articles have been featured in AARP The Magazine, Reader’s Digest, The Costco Connection, and other publications. He often writes on topics related to real estate, business, technology, healthcare, insurance, and entertainment. He also publishes several blogs, including martinspiration.com and cineversegroup.com.