GPT-4o set to disrupt conversational AI, edtech and call centre business (2024)

OpenAI’s latest GPT-4o (“omni”) model has not just ‘wowed’ people across the globe but is set to disrupt everything from conversational AI and edtech to content creation and customer services, experts said.

GPT-4o, launched on Monday, will also majorly impact contact centres business, propelling towards fully autonomous call centres without any dependency on human agents, they said.

“It seems OpenAI has built an audio foundation model ground up – the model can sing, change tones, speak slow or fast, understand your emotions and respond appropriately,” said Hemant Mohapatra, partner at venture capital firm Lightspeed.

ETtech

“A full stack voice and workflows that GPT offers out-of-the box very cheap will be very attractive to developers,” he said, adding that business applications built on caller APIs will find it tough to justify applications.

Voice-first AI startups such as ElevenLabs, play.ht, Respeecher, Hume and Duolingo are bracing up for increased competition from GPT’s multimodal capabilities and real-time language translations.

In fact, after the GPT-4o demo showed real-time language translation, Duolingo lost $340 million in market cap on Monday.

“We are moving closer to AGI (artificial general intelligence) faster than we are willing to believe,” said Alok Goyal, partner at venture capital firm Stellaris Venture Partners. “The only bad news of course is that humans will soon need to worry about their own role and expertise!”

With GPT4o, he said, many complex, cognitive tasks can be done “off the shelf” with no fine-tuning/training in areas such as education, healthcare, financial services, and sales.

Kunal Bahl, cofounder of venture capital firm Titan Capital, said GPT-4o is “the ultimate edtech utility” that could teach millions of students in their native language who wouldn’t need teachers or schools.

The GPT-4o model will power its free-to-use chat assistant ChatGPT with multimodal capabilities across text, vision and audio.

The ‘omni’ model can reason across formats in real time and perform tasks such as debugging code via voice commands, make jokes, sing a lullaby, synthesise 3D objects, resolve customer service complaints, and even talk to another AI.

Raghu Ravinutala, CEO and cofounder of conversational AI startup Yellow.ai, said GPT-4o brings us closer to AI agents very close to humans. “Imagine an AI-powered voice that understands and adapts to your frustration in a conversation, changing its tonality accordingly,” he said.

“The previous voice mode used multiple steps: converting audio to text, processing it through a language model, and then converting the text back to speech,” Ravinutala said. “GPT-4o simplifies this by managing everything within a single model, significantly reducing latency and enabling real-time interpretation for conversations in distinct languages.”

Further, GPT-4o is poised to enhance accessibility for visually impaired individuals, as demonstrated in OpenAI’s videos.

However, the transformation it is expected to bring will not be easy and won’t happen overnight, experts said.

“Just think for a second when you want to change a flight… There are so many variables in play such as price, time, airline, duration, etc.,” said Aakrit Vaish, co-founder and CEO of Indian conversational AI platform Haptik. “Often, these are better done through GUI or a hybrid interface. AI will get some things wrong, just like humans will. But, with AI, we blame the person who built the AI (brand) and not say GPT,” he added.

Amitabh Nag, chief executive of Bhashini, a Digital India initiative to make digital services accessible in all Indian languages, said, “Our aim is to reach India’s last mile, i.e., people living on the other side of not only language divide but also digital divide and literacy divide. To achieve that, AI knowledge, glossary, etc. need to be Indianised to a large extent, and it remains to be seen if global AI models can achieve that.

Experts said AI-native startups will expand their margins as OpenAI has cut the price of business API into half.

Stellaris’ Goyal also reckoned that hyperscalers will witness a boom as “media-rich applications continue to require even more compute and storage”.

AI video analytics company Staqu Technologies said cost of computation is going to be a major factor in full-scale adoption. “For a street camera installed in a smart city and monitoring large crowds 24/7, AI will run a huge bill in terms of both computational power and cost,” Atul Rai, founder of Staqu, said.

Pawan Prabhat, cofounder of generative AI firm Shorthills AI, said Indians are mostly voice-first technology consumers and GPT-4o will bring a quantum jump in AI consumption among India.

“Till now, most of the interaction has been through keyboard and typing. As humans, we are wired to ‘show and tell’. GPT-4o gets AI closer to the way humans converse and interact,” he said.