OpenAI takes its widely criticized AI-written text detector off the market due to an unsettling ‘low rate of accuracy’, shedding light on the difficulties of reliably identifying AI-generated text.


OpenAI scuttles AI-written text detector over ‘low rate of accuracy’

OpenAI’s pulled the plug on its AI detector tool that was supposed to figure out if a piece of text was written by an AI. Turns out, the tool was not hitting the mark most of the time. The idea that AI-generated text has certain giveaway signs sounds good, but in reality, it hasn’t been proven to be foolproof. TechCrunch put several of these AI-spotting tools through their paces and found them to be dodgy at best. Of the seven text bits they used, OpenAI’s tool could only correctly flag one.

Even though OpenAI warned folks about the limitations of their tool, some still used it and placed a bit too much faith in the results. The folks at OpenAI seem to have finally decided to retire this unpredictable tool, stating they’re now focused on finding better ways to trace the origin of text.

Interestingly, this move came around the same time OpenAI and some other companies joined a “voluntary commitment” led by the White House to develop AI in an ethical and transparent way, which includes creating reliable ways to track and/or detect AI-generated content. Yet, so far, there’s no foolproof solution in sight. Whoever cracks this nut first is likely to strike gold, since a dependable tool would be a game-changer in many situations.


Driven by AI boom, TSMC to invest $2.9 billion in advanced chip plant in Taiwan

TSMC, a Taiwanese chipmaker, plans to shell out $2.87 billion for a new advanced chip facility in Taiwan. The boom in AI has skyrocketed the demand for their chips and they can’t keep up. To help meet the demand, TSMC is going to double their advanced packaging efforts, which is a method of squeezing multiple chips into one device to boost computing power without hiking up the price too much.

Their advanced packaging, particularly chip on wafer on substrate (CoWoS), is in high demand. TSMC’s CEO, C.C. Wei, said they are cranking up their capacity as fast as they can and expect to be able to breathe easier by the end of next year.

They’ve got the green light to lease land in the Tongluo Science Park, and the new facility will create around 1500 jobs. While TSMC is branching out, it intends to keep its cutting-edge chip tech in Taiwan, a major player in making semiconductors that power everything from smartphones to electric cars.


You can now chat with a famous AI character on Viber

Messaging app Rakuten Viber has announced a partnership with ICONIQ to bring the AI character Kuki onto its platform. Unlike typical AI assistants, Kuki, known for modeling for Vogue and being a brand ambassador for H&M, is designed to entertain users rather than assist with tasks.

Kuki can participate in open-ended conversations, providing companionship and entertainment, and is available for a variety of activities including tarot card readings, playing games, providing horoscopes, or simply chatting. Users can access Kuki at any time by searching in the chat function of the Viber app or on its explore page, and then subscribing to the chatbot.

This move reflects a growing trend in the messaging app industry to integrate AI to enhance user experience and provide unique entertainment opportunities. According to Ofir Eyal, CEO of Rakuten Viber, AI companions like Kuki offer users an entirely unique entertainment experience.


Meet DreamBooth: An AI Technique For Subject-Driven Text-to-Image Generation

A novel AI approach named DreamBooth has been introduced that allows for a more personalized experience in text-to-image diffusion models. This methodology can help generate images that are more in line with individual users’ unique requirements. By expanding the model’s language-vision dictionary, the new approach helps to establish associations between new words and specific subjects users want to generate.

While large-scale text-to-image models can generate high-quality images based on language descriptions, they often fail to faithfully replicate the appearance of subjects from a reference set or to generate new interpretations of those subjects in different contexts. The introduction of DreamBooth addresses this limitation, allowing models to synthesize photorealistic images of the subject set within different scenes while preserving their distinctive identifying features.

DreamBooth embeds the subject into the model’s output domain in a manner that allows its synthesis with a unique identifier, using a small collection of subject images (around 3-5). The technique employs rare token identifiers and fine-tuning of a pre-trained, diffusion-based text-to-image framework.


AI Tool Creates South Park Episodes With User In Starring Role

US-based Fable Simulation has developed an artificial intelligence (AI) tool, AI Showrunner, which can produce short original episodes of the animated series South Park, featuring users as the main characters. The tool, developed without copyright permission and only for research purposes, can create episodes based on one- or two-sentence prompts. It is capable of animation, voice generation, and editing.

The technology, known as generative AI, has attracted public attention since the launch of ChatGPT, and it has led to discussions about its potential role in Hollywood. Edward Saatchi, CEO of Fable Simulation, emphasized that they don’t intend to profit from the tool or make it publicly available. They used South Park as a research example to demonstrate how the model works.

Interestingly, South Park’s creators, Trey Parker and Matt Stone, have not been approached regarding the experiment, despite their own interest in AI. Last year, they secured a $20m investment for their AI entertainment company, Deep Voodoo.


At Startup That Says Its AI Writes Medical Records, Humans Do a Lot of the Work

The health-tech startup DeepScribe has boasted about its AI’s power to convert doc-patient conversations into accurate medical records. However, it’s not as autonomous as advertised. Many slip-ups, including incorrect medical terms and throwing in medicines a patient isn’t popping, have required the company to rope in about 200 real folks to tidy up the records. They even resort to Googling for billing codes. DeepScribe’s head honchos admit their AI can handle 80% of the job but credit their human backup for catching errors and bolstering trust in the results.

Doctors, who are the final gatekeepers of these records, are reportedly catching the mistakes that manage to slip through the human review. The use of AI in healthcare record-keeping can potentially save doctors from the drudgery of note-taking, which is critical for tracking a patient’s health history and deciding on treatment. The tedious task usually involves copying and tweaking previous notes, dictating reports, or having a staff member note down the conversation.

Despite the issues, DeepScribe remains confident in its tech. They stress that their AI doesn’t just transcribe notes but turns lengthy chats into standardized reports that are easily integrated into patient records.

READ THE ARTICLE ON THE WALL STREET JOURNAL Raised A $5 Million Pre-seed Round With This 11-slide Pitch Deck, a fresh AI startup focused on professional upskilling, has scored $5 million in pre-seed funding. The company, brainchild of Onfido cofounder Husayn Kassai, aims to help folks find the right learning materials to enhance their skills.

Quench has created an AI learning coach to help workers make the most out of their company’s resources. In Kassai’s previous experience, the top learners were often the top performers. He believes AI can help folks find the training they need to level up their skills and move faster.

Quench’s AI can answer a bunch of questions usually asked by newer employees. With more people working remotely due to the pandemic, many miss out on informal learning opportunities, like office chit-chat. Kassai says AI offers a new approach to learning, which is critical for startups where speed and learning ability can make or break success.


Citi stays positive on A.I. theme and lays out the key to finding winning stocks

Citigroup remains optimistic about the future of artificial intelligence (AI), and it believes that cash flows will be key to identifying the winning stocks in the next phase of AI investment. According to Citigroup, medium to long-term opportunities still exist within the AI sector due to its accelerating growth trajectory and appealing free cash flow (FCF) dynamics.

AI-related stocks have seen a significant increase in valuation so far this year, with Nvidia shares surging more than 200%. Citigroup maintains that the initial positive thesis is intact and warns investors not to overlook the importance of free cash flows. The bank sees profitable stocks within the AI sector as significant cash generators, and it believes recent developments in AI will push FCF margins and growth to new highs.

In a recent screening for AI-related stocks expected to outpace market growth expectations and experience a surge in free cash flow margins, Amazon emerged as a leader. Amazon is expected to grow by more than 48% over the long term, and its shares have already gained nearly 54% this year. Although some investors consider Amazon to be lagging behind in the AI race, CEO Andy Jassy recently reassured investors by reaffirming the company’s commitment to investing in AI across various segments.


Growth driven by AI will be the metric to watch this earnings cycle

Investments in AI technology by major tech companies such as Alphabet and Microsoft are becoming a key factor that investors are looking at this earnings season. Both companies have benefitted from their early forays into the AI sector, with Microsoft’s share prices increasing around 44% so far this year and Alphabet’s stock rising by 36%.

The growth of large language models (LLMs), which power generative AI services, has also drawn significant attention from investors. As a result, tech companies’ prominence in the AI field has become a new benchmark for evaluating their potential for growth and innovation. Demonstrations of any advancement in AI technology can cause a significant increase in share prices.

Other tech giants, such as Meta, have also experienced impressive growth this year, with Meta’s shares increasing by a striking 134%. However, despite the rally in tech shares, there are concerns about a potential recession and the end of a prosperous period in the tech sector.


Video game studios are so scared of A.I. they’re forcing managers to study machine learning and offering employees $7,000 bounties for A.I. ideas

Video game studios worldwide are in a race to integrate and develop AI tools in order to cope with the skyrocketing costs of game production. In this scramble, companies are offering significant incentives, such as bonuses of up to $7,000 for innovative AI ideas, to keep pace with the industry’s rapid AI transformation.

AI has presented the gaming industry with an opportunity to revise bloated and formulaic business models. High production costs have been a key concern, with blockbuster games often costing in excess of $200 million. AI could potentially halve the time and cost of these projects.

The integration of AI in the gaming industry might bring about a significant shift, where the value of games is determined by creativity rather than budget. While this transition period might be painful for some, it could potentially benefit gamers around the world by diversifying the types of games available and enabling smaller studios to produce high-quality games at a fraction of the cost.