• AI KATANA
  • Posts
  • How OpenAI designed the ChatGPT app to be addictively simple

How OpenAI designed the ChatGPT app to be addictively simple

Also: More people in Korea are using AI-edited photos as their headshots on their resumes

Author

AI KATANA
June 23, 2023

Hello!

Welcome to AI KATANA. OpenAI's ChatGPT app achieved an impressive milestone, reaching 100 million active users within two months, thanks to its simplistic, intuitive design. Meanwhile, Forrester Research predicts that by 2030, advertising agencies could automate 7.5% of jobs due to generative AI.

Elsewhere, Hugging Face's CEO testified before the U.S. House Science Committee, emphasizing the crucial role of open-source AI in innovation and its alignment with American interests. Additionally, new AI tools are emerging, like Inflection's language model, Aloud's AI-powered dubbing for YouTube, and Voicemod's real-time voice modification for gamers. Lastly, AWS invested $100 million in generative AI, marking a significant move in the AI industry.

Anyone want a slice?

  • 📱 How Open AI designed the ChatGPT app to be addictively simple

  • 👨 Agencies to replace 7.5% of jobs with AI by 2030

  • 🇺🇸 Hugging Face CEO tells US House open-source AI is 'extremely aligned' with American interests

  • 🇰🇷 More people in Korea are using AI-edited photos as their headshots on their resumes

The article discusses how OpenAI's user experience team, led by Patrik Goethe, designed the ChatGPT app for iOS to be extremely simple and intuitive. Guided by the mantra to "kill any complications, avoid distractions, and keep it simple," the team developed an interface that is minimalist and chat-centric, resembling an old version of AOL Instant Messenger. The team believed text to be the most powerful form of communication between a human and a computer and thus focused the design around a simple action: users type or voice-upload any prompt they can think of, and the AI responds with lines of text. The article also notes that the success of the app, which reached 100 million monthly active users just two months after its November 2022 launch, is based on the power and universality of language. The design aligns with the evolution of user interfaces, from text-based command line interfaces to graphical user interfaces, and now back to text again with AI integration. This makes the interface more organic and intuitive, leading to a less steep learning curve than is typically associated with advanced technology.

According to a forecast by Forrester Research, advertising agencies are set to replace 7.5% of jobs with automation by 2030 due to the rise of generative AI. This will put about one-third of agency jobs at risk, with roles in clerical, secretarial, sales, and market research most threatened, amounting to 28%, 22%, and 18% of job losses respectively. Despite the apprehension, creative roles such as writers and editors are seen to be among the least at risk. Generative AI is expected to enhance productivity in these areas, potentially leading to more capable agencies. Already, 56% of B2C marketers are using generative AI, with an additional 40% in the exploratory phase. Several companies and agencies, including Coca-Cola, WPP, and Nvidia, are actively leveraging generative AI. The impact of generative AI on agency jobs over the next two years is predicted to be modest due to ethical and legal considerations, but marketers are advised to start integrating AI to avoid falling behind.

Hugging Face CEO, Clement Delangue, testified before the U.S. House Science Committee, asserting that open science and open-source AI are essential for innovation and strongly aligned with American values and interests. Delangue stated that open source initiatives have fueled the progress of AI, with key technologies such as PyTorch, Tensorflow, Keras, transformers, and diffusers—all of which were invented in the U.S.—contributing to America's leading position in AI. His testimony follows concerns raised by Senators about the potential misuse of open-source AI models like Meta's LLaMA. Delangue argued that open science and open source foster startup creation, increase accountability, mitigate biases, and combat misinformation. He also highlighted Hugging Face's approach to ethical openness, which includes documentation, technical safeguards, and community incentives. Delangue's comments occur amidst a broader debate on the shift from open-source AI to proprietary large language models.

In South Korea, the Millennial and Gen Z generations are showing increased interest in the 'AI Profile' service by SNOW, a subsidiary of Naver. This service creates high-quality AI-generated photos that resemble professional studio pictures, offering an affordable alternative to traditional photography. It has gained tremendous popularity since its launch on May 25. However, despite the increasing trend of using these AI-edited photos as profile pictures on resumes, experts advise against it. Representatives from the employment platform Job Korea and the education industry emphasize the importance of clear, professionally taken photos that best represent the individual's impression. A survey conducted by Job Korea indicated that 58.2% of hiring managers value the 'impression' conveyed by a photo more than appearance, suggesting that excessively edited photos could hinder interview performance. Instead, experts recommend subtle alterations like balancing lighting and cleaning up skin tone.

🛠️ AI tools update

AI startup Inflection has unveiled a new large language model, Inflection-1, which powers its Pi conversational agent. The model is of roughly GPT-3.5 size and capabilities, as measured by the computing power used to train it. The company claims it's competitive or even superior to other models on this tier, and supports these claims with a "technical memo" describing some benchmarks it ran on its model and others. The results indicate that Inflection-1 performs well on various measures, such as middle and high school-level exam tasks and "common sense" benchmarks, though it lags behind in coding tasks. Inflection plans to publish results for a larger model comparable to GPT-4 and PaLM-2(L) in the future. However, the effectiveness of these models will ultimately be determined by their widespread use and independent evaluation. Inflection-1, as part of the Pi conversational agent, is currently available for testing via messaging apps or online.

YouTube has announced its partnership with Aloud, an AI-powered dubbing service from Google's Area 120 incubator, to make it easier for content creators to dub their videos in multiple languages. The AI tool works by first transcribing the video, allowing for review and edits, before translating and producing the dubbed content. Currently under testing with hundreds of creators, Aloud supports a few languages, including English, Spanish, and Portuguese, with plans to add more. Despite its current language limitations, Aloud proves to be a valuable resource as an increasing number of creators are adding multi-language dubs to their videos. Looking ahead, YouTube plans to enhance the translated audio tracks to more closely resemble the creator's voice in terms of expression and lip sync, with these features expected to roll out in 2024.

Valencia-based artificial intelligence firm Voicemod has unveiled an array of "AI humans" that enable users to modify their voices in real-time while playing video games. The technology, which allows users to switch between 20 different digital personas, from a high-pitched female speaker to a baritone male voice, is used as a "virtual microphone" that sits between the user's microphone application. With a user base of over 40 million people, Voicemod's technology is particularly popular among the social app Discord's community, with CEO Jaime Bosch highlighting its positive impact on shy users who find it easier to socialize through these alternative personas. This innovation is a milestone in conversational AI and requires a significant amount of computing power, investment, and talent to ensure real-time voice modification. In light of potential misuse, Voicemod is developing a watermarking solution to identify voices generated using modification systems and is engaged in discussions about standardizing such systems. The company has raised $23 million in funding from venture capitalists, including Leadwind and Bitkraft Ventures.

💵 Venture Capital updates

Amazon Web Services (AWS), Amazon's cloud unit, has announced an investment of $100 million in a program designed to assist companies in leveraging generative artificial intelligence (AI), a field that has gained momentum since OpenAI launched its ChatGPT chatbot. This investment, though relatively small for a company of Amazon's size, underscores AWS's recognition of the importance of generative AI and its desire to remain competitive with rivals Microsoft and Google, who have made notable strides in this area. As part of this initiative, AWS will be hiring additional data scientists, engineers, and solutions architects, and the program is already working with companies like Highspot, Twilio, RyanAir, and Lonely Planet. AWS CEO Adam Selipsky has likened the current state of the AI race to the early stages of a long-distance run, implying that it's still early days for this technology.

Paris-based AI startup Mistral recently raised €105m just four weeks after its launch, without a product, causing divided responses in the tech community. Some view it as a significant achievement for European tech, with talent from Meta and Google founding a potential rival to OpenAI, while others see it as an example of the AI hype leading to large investment rounds for inexperienced founders. The startup's pitch deck, or more accurately, a Google doc, played a crucial role in convincing investors of its potential. The founders, Timothée Lacroix, Guillaume Lample, and Arthur Mensch, leveraged their deep experience in AI and machine learning at Meta and Google's DeepMind to argue the value of large language models (LLMs), the hard-to-make technology at the heart of the emerging generative AI market. They believe that the value in this sector will come from the creators of the models themselves and that Europe lacks serious contenders in building these foundational models.

🫡 Meme of the day

⭐️ Generative AI image of the day