From sci-fi to reality: The dawn of emotionally intelligent AI

Sometimes it feels as if the world suddenly jumps forward. Certainly, that’s what it felt like 18 months ago when OpenAI first released ChatGPT. According to a study from UBS, ChatGPT grew to more than 100 million users in the first two months of its release, making it the fastest-growing consumer application in history. At the time, there was a palpable sense that something dramatic was happening.

Then the world careened from fascination and excitement to fear and anxiety about the potential for imminent extermination from rogue AIs. Remember the “pause” proposed by Elon Musk and several hundred others, arguing that AI development should stop for six months? This didn’t happen, though Musk — ironically — began building his own ChatGPT-like alternative during this time.

Some, such as AI researcher Eliezer Yudkowsky, worried that ChatGPT pointed towards a near future when AI “gets to smarter-than-human intelligence” and kills everyone. It even became fashionable for people to assess their belief in the likelihood of this outcome, referred to as the “probability of doom” or “P(doom).” “What’s your P-(doom)?” was the subject of dinner conversations, at least in Silicon Valley and other high tech hangouts.

People who claimed a high P(doom) were labeled as “doomers.” This even included Geoffrey Hinton, one of the “Godfathers of AI,” who decided a year ago that the AI technology he had helped to advance was advancing much faster than he expected. His concerns over potential existential risks from increasingly powerful AI systems prompted his change in stance.

Paradigm shift

Nevertheless, we are still here. Perhaps the risks are still here too — just existing sometime in the future. But instead of The Terminator, we have gotten friendly virtual assistants like Samantha. This new assistant capability is based on GPT-4o (“o” for omni), released this week from OpenAI, that can “reason across audio, vision, and text in real time.” It can converse using natural-sounding speech without noticeable lag time, read emotional cues (and respond in kind), and process visual input. In line with these new capabilities, senior writer Mark Sullivan at Fast Company concluded that ChatGPT “is getting more emotionally intelligent.”

AI Editor Ryan Morrison at Tom’s Guide posited: “What we’re witnessing — and other companies will catch up — is the dawn of a new era in human-computer interface technology.” Morrison added: This “is as big of a paradigm shift in accessing information as the first printing press, the typewriter, the personal computer, the internet or even the smartphone.” I had to agree with this observation after watching the live demo of GPT-4o. It indeed felt like another significant jump ahead.

AGI or narrow AI?

And yet, this really should not be happening, at least not now. Only a few months ago, it was believed that this level of achievement would require artificial general intelligence (AGI), requiring  significant scientific advances in areas such as natural language processing, emotional intelligence, and creativity.

No one is claiming that this new level of virtual assistant is AGI, a general intelligence that matches or exceeds human-level performance across a wide range of cognitive tasks. I asked ChatGPT about this and it agreed that GPT-4o is not yet AGI. It said: “GPT-4o should be described as an advanced form of weak [or narrow] AI. It’s a sophisticated tool that blends multiple types of data processing in innovative ways, pushing the boundaries of what AI can do within specified parameters but still operating under the limitations of a machine designed for particular applications, rather than a fully autonomous, general-purpose intellect.”

Parlor trick or useful for applications?

The GPT-4o demo felt like wizardry, but will this translate into real world usefulness? Adding human-like interaction will make AI technologies feel more natural and engaging for the user and could clearly enhance user experiences across various applications — including customer service, virtual assistance, education and entertainment.

Kevin Roose, a tech columnist at The New York Times, recently talked on the Hard Fork podcast and in the Times about the “AI friends” he created across several platforms including Nomi, Kindroid and Replika. All these companies use large language model (LLM) technology — like what underpins ChatGPT prior to GPT-4o. Roose built 18 personalized companions, gave each a backstory, and was able to chat with them texting back and forth. He could talk with some of them too, though with a substantial lag time. Even before the news of GPT-4o and its ability to “see” and communicate emotions, and with voice responses close to real-time, he concluded: “The technology needed for realistic AI companionship is already here.”

Other companies are already building on top of GPT-4o, including one that is adding a face to the voice. With synchronized expressions and lip movement, this capability will add to the already notable appeal of AI friends, but will also find application for healthcare, caregiving, coaching and education — where a friendly face can make a difference. The addition of a human-like face makes it easier to imagine virtual teachers, healthcare advisors or brand representatives.

Greater ability to integrate audio, visual and textual data opens potentially transformative opportunities in sectors like healthcare, where it could lead to more nuanced patient interaction and personalized care plans.

Doctors are already exploring the possible impact of this advance, including virtual health assistants operating as “24/7 health advisors, providing immediate medical advice, reminding patients about medications and even offering mental health support.” This could also include “empathetic support for mental health issues,” that supplement traditional therapies by recognizing and responding to subtle cues in a patient’s speech or behavior.

In education, this technology could cater to varied learning styles with adaptive content that responds in real-time to the needs of each student. For example, in this video, Sal Kahn — the founder of the Khan Academy whose mission is providing free education — and his son Imran work with GPT-4o on solving a math problem. 

As GPT-4o and similar offerings continue to evolve, we can anticipate more sophisticated forms of natural language understanding and emotional intelligence. This could lead to AI that not only understands complex human emotions but also responds in increasingly appropriate and helpful ways. The future might see AI becoming an integral part of emotional support networks, providing companionship and aid that feels genuinely empathetic and informed.

The journey of AI from niche technology to a fundamental part of our daily interactions is both exhilarating and daunting. To navigate this AI revolution responsibly, it is essential for developers, users and policymakers to engage in a rigorous and ongoing dialogue about the ethical use of these technologies. As GPT-4o and similar AI tools become more embedded in our daily lives, we must navigate this transformative journey with wisdom and foresight, ensuring AI remains a tool that empowers rather than diminishes our humanity.

Gary Grossman is EVP of technology practice at Edelman and global lead of the Edelman AI Center of Excellence.