Voice AI has been around for quite a while, but it hasn’t always been smooth sailing. Early versions often sounded robotic, had awkward pauses, and sometimes got confused, which kept many companies from fully adopting the technology. However, recent talks at the Goldman Sachs Communacopia + Technology conference suggest that these issues are starting to fade as new innovations and more startups jump into the voice AI scene.

Why Voice AI Is Getting Better

Twilio and Zoom CEOs shared their insights on how problems like slow responses and unnatural voices are being addressed. Twilio’s CEO, Khozema Shipchandler, mentioned that internal studies show many customers actually prefer talking to voice AI instead of humans, especially in healthcare. People feel there’s an imbalance of knowledge when dealing with human agents, which can make interactions feel awkward. Virtual voice agents help remove those uncomfortable moments and make conversations smoother.

Shipchandler also pointed out that latency, or the delay in responses, has been a big hurdle. But now, that delay is becoming less noticeable. Zoom has invested heavily in its voice AI, which can handle multiple languages and produce more natural-sounding voices. The aim is to eliminate those strange pauses during calls. Yet, real-world tests have shown mixed results. Some fast-food chains like Taco Bell and McDonald’s have even stopped using voice AI at drive-thrus because it couldn’t understand orders properly. The tech is more challenging to get right than text-based AI, according to Jack Gold, a tech analyst. Variations in accents and dialects make it tricky for AI to interpret speech accurately. Still, voice remains a natural way for many to communicate, especially for those who find typing difficult. For instance, around 35% of food delivery orders still come in over the phone, and voice AI can help speed up those calls.

The Future of Voice AI and Its Challenges

Shipchandler believes that voice AI’s capabilities are virtually unlimited, and thousands of startups are working to solve its current problems. Zoom CEO Eric Yuan also noted that more people are now talking to ChatGPT instead of just typing prompts. He predicts that over the next two or three years, a lot of new voice-based solutions will be developed.

Despite the progress, there are still safety concerns like voice spoofing, where bad actors mimic someone’s voice to trick systems. To combat this, experts suggest systems could use voice signatures to verify identities early in the interaction, making conversations more secure and seamless. Shipchandler emphasized the importance of eliminating spoofing because it’s a real threat.

Meanwhile, Zoom is collaborating with security experts and publishing guidelines on how to safely deploy voice AI. Gold expects ongoing improvements in the technology, especially as the data fed into these systems gets better. The more accurate and diverse the data, the more reliable the AI becomes in understanding speech.

What’s Next for Voice AI

As voice AI continues to improve, it’s expected to become a key part of many industries. Gold points out that the errors we see now will decrease as the models learn from better data and new innovations. The potential is huge, from customer service to healthcare, where clear, quick voice communication can make a big difference.

Yuan from Zoom believes that in the next few years, voice will be the foundation for many new tech solutions. While there are still hurdles like security and accuracy, the momentum is clear. The tech world is betting big on voice AI becoming more natural, reliable, and widespread. So, it looks like the future of voice AI is bright — with fewer awkward pauses and more meaningful conversations.

Inspired by

• https://www.computerworld.com/article/4073954/voice-ai-issues-the-clunky-speech-weird-pauses-and-inaccuracies-are-being-fixed-execs-say.html

Sources

• gmpg.org
• bbc.com
• wsj.com
• spglobal.com
• theregister.com