OpenAI has introduced the latest iteration of the technology powering its AI chatbot ChatGPT.
Dubbed GPT-4o, this version will be made available to all ChatGPT users, including those who are not subscribers.
GPT-4o boasts improved speed compared to its predecessors and has been designed to offer more conversational and even flirtatious responses to prompts.
Additionally, this new version is equipped with the ability to analyze and discuss images, translate languages, and recognize emotions based on visual cues. It also features a memory function, enabling it to recall previous prompts.
Moreover, GPT-4o exhibits a smoother conversational flow, with minimal delay between questions and responses, and it can be interrupted during interactions.
Glitches
In a live demonstration featuring the voice version of GPT-4o, the AI provided useful guidance on approaching the solution of a basic equation written on paper, rather than simply delivering the answer. It also analyzed computer code, facilitating translation between Italian and English, and interpreted emotions in a selfie of a smiling man.
Delivered in a friendly American female voice, GPT-4o greeted its users by inquiring about their well-being. Upon receiving a compliment, its response was: “Stop it, you’re making me blush!”.
It exhibited imperfections – at one instance, it mistook the smiling man for a wooden surface, and it attempted to solve an equation it hadn’t been presented with yet. This inadvertently underscored that there’s still progress needed to address glitches and hallucinations that render chatbots unreliable and potentially risky.
However, what it does indicate is the trajectory for OpenAI, which seems to envision GPT-4o as the next evolution of AI digital assistants, akin to a highly advanced Siri or Hey, Google that retains memory of past interactions and can engage beyond voice or text.
If there was a looming concern, amidst the enthusiastic off-camera audience cheering and applause, it was the environmental footprint of this technology.
We’re aware that AI demands more power than traditional computing tasks, and as it becomes more intricate, its computational demands increase. Sustainability was conspicuously absent from the discussion during the event.
Demystify
We’ve witnessed chatbots such as Elon Musk’s Grok and Pi, developed by DeepMind co-founder Mustafa Suleyman, prioritize the “personality” of their offerings. However, the seamless integration of text, audio, and images with instant responses showcased by GPT-4o appears to position OpenAI at the forefront of the competition.
Nevertheless, at this juncture, we solely rely on the company’s assurances – it was their demonstration, meticulously crafted and orchestrated by them. It will be intriguing to observe how GPT-4o handles the influx of users as it is rolled out on a larger scale, especially among the millions who already utilize ChatGPT.
OpenAI’s Chief Technology Officer, Mira Murati, described GPT-4o as “magical” but added that the firm would “remove that mysticism” with the product’s roll-out.
A captivating and evocative selection of words: although this technology is swiftly advancing in sophistication and becoming more persuasive as a companion, it remains devoid of sentience or magical properties; it is the result of intricate programming and machine learning.
Speculations have surfaced regarding a potential collaboration between OpenAI and Apple, although this has yet to be officially confirmed. However, it was notable during the demonstration that Apple products were prominently featured.
Furthermore, the timing of this event, occurring just 24 hours prior to its competitor Google’s annual conference, Google IO, serves as a strategic maneuver and indicates a subtle challenge from OpenAI.