Welcome to the latest issue of Hunt Good Weekly!
In this issue you will see:
5 fresh news
4 useful tools
3 distinct points of view
1 interesting case
Hunt for News｜Advanced Headlines
A prophet emerges: 10 predictions for the next year in the AI industry
According to X blogger @op7418, the well-known venture capital firm Air Street Capital recently released the 2023 Artificial Intelligence Status Report, covering all aspects of artificial intelligence research, industry, security and politics.
Every year, Air Street Capital makes 10 predictions about the AI industry for the next year. Among the 10 predictions released last year, 5 predictions were accurate.
Here are 10 predictions for next year
- Hollywood has begun to use generative AI to produce visual effects, such as Marvel, which has begun to experiment.
- A generative AI media company is under investigation due to misuse during the 2024 US election.
- AI agents that can improve themselves exceed industry standards in complex environments, such as tool usage and scientific research.
- The IPO market in the AI field is starting to recover, and we're seeing at least one AI-focused company go public.
- The craze for generative AI has led one team to spend more than $1 billion training a large model.
- The FTC or CMA is investigating the deal between Microsoft and OpenAI on competitive grounds.
- Progress in global AI governance has been limited, consisting of high-level voluntary commitments.
- Financial institutions launch GPU debt funds to provide financing for computing as an alternative to venture capital.
- An AI-generated song enters the Billboard top 10 or a Spotify top song of 2024.
- A large AI company will acquire a chip company due to a significant increase in costs.
Which prediction do you think will come true? Please share your opinion in the comment area.
Employees voted for the most popular CEO, and Jen-Hsun Huang topped the list
The most successful companies are not necessarily led by the most charismatic CEOs. According to Blind, an anonymous discussion community for technology employees, 13,171 U.S. employees voted on 103 CEOs:
Do you approve or disapprove of the way your CEO handles his job?
The results show that Jen-Hsun Huang is far ahead with an approval rating of 96%, and the second-ranked CEO is Doug McMillon from Walmart, which is a full 8 percentage points lower than Jen-Hsun Huang.
Apple CEO Cook ranked fourth with an approval rating of 83%. He is one of the few leaders of large technology companies to receive enthusiastic support from employees. In comparison, Google CEO Pichai only received 26%.
The two CEOs at the bottom are Western Digital's David Goeckeler and Nordstrom's Erik Nordstrom.
In June this year, Western Digital laid off more than 200 employees. Interestingly, according to the company's fiscal 2022 summary, Goeckler's income is 3,332 times that of the average Western Digital employee.
OpenAI plans a major update that could reduce developer costs by 20 times
OpenAI plans to roll out a major update for developers next month that will make it cheaper and faster to build software applications based on its artificial intelligence models, sources told Reuters.
The updates include adding in-memory storage capabilities to its development tools for working with artificial intelligence models. In theory, this will reduce costs for application manufacturers by up to 20 times.
Businesses can create applications more cheaply by remembering conversations for queries, which can significantly reduce the amount of usage developers need to pay for.
Currently, you can pay 10 cents to process a page of a document using GPT-4, depending on the length and complexity of the input and output, according to pricing on the OpenAI website.
In addition, OpenAI also plans to launch new tools such as vision functions to enable developers to build applications with the ability to analyze and describe images. Potential application scenarios range from entertainment to medicine.
The new features are expected to be launched at OpenAI's first developer conference in San Francisco on November 6, and are designed to encourage companies to use OpenAI's technology to build artificial intelligence-driven chatbots and autonomous agents, people familiar with the matter said.
CMA formally approves Microsoft's acquisition of Activision Blizzard
Britain's antitrust regulator, the CMA, yesterday formally approved Microsoft's $68.7 billion acquisition of Activision Blizzard, removing the last major obstacle to completion of the deal.
The CMA said it had approved Microsoft's deal to acquire Activision Blizzard, but not the cloud gaming rights. The CMA said that Microsoft made concessions in August this year, and Ubisoft will purchase Activision Blizzard's cloud gaming rights.
Microsoft President Brad Smith said excitedly on the X platform:
"We thank the CMA for today's thorough review and decision. We have now crossed the final regulatory hurdle and completed this acquisition, which we believe will benefit players and the gaming industry around the world."
In January 2022, Microsoft announced its acquisition of Activision Blizzard for US$68.7 billion, setting an acquisition record in the gaming industry.
"AI Godfather" master and apprentice join forces to enter robot entrepreneurship
Turing Award winner and the father of deep learning, Geoffrey Hinton, recently announced that he will join the robotics startup Vayu Robotics as a consultant.
In May of this year, according to the New York Times, Geoffrey Hinton, the "Godfather of AI" and Turing Award winner, left Google, where he had served for ten years, in order to be able to freely discuss the risks of artificial intelligence.
When talking about the reasons for joining the startup Vayu Robotics, Hinton said that their technical route has lower moral hazard than many other AI applications.
Nitish Srivastava, CTO of Vayu Robotics, is Hinton’s disciple. Therefore, Hinton's joining this time is more like a joint venture between teachers and students.
Hunt for Tools｜Advanced Tools
Is Google AI search equipped with MIdjourney? Easy Wensheng Pictures
At the Search On event on October 12, Google announced the withdrawal of new AI search features.
For example, when you can't find a suitable picture using a search engine, you can use descriptive prompts to specify the image you want to generate.
This feature is powered by Google Images' image generation capabilities, with up to four generated images available in the results. In other words, Google has integrated image generation solutions like DALL-E or MIdjourney into traditional search engines.
Hema Budaraju, senior director of search product management, said each image generated with SGE will be "tagged with metadata and embedded with a watermark to indicate that it was created by AI."
In addition, based on the latest writing draft function, you can let it automatically output long articles of different lengths and styles. It’s worth noting that the update is currently only available in the United States and will be rolled out to more regions in the coming weeks.
Previously, we have also had detailed tracking reports on Google’s AI-generated search. Click on the title to jump to read .
Adobe releases new generation AI generation model Firefly Image 2
At the annual MAX Creative Conference, Adobe launched three AI models: Adobe Firefly Image 2, Adobe Firefly Vector Model and Adobe Firefly Design Model, which are the latest versions of its AIGC mapping platform Firefly.
Among them, Firefly Image 2 supports popular functions such as Photoshop's generated fill, and also adds a new vector generation function, which can directly generate vector graphics through prompt.
Adobe says the Firefly Image 2 model is capable of producing higher-quality images than its predecessor, especially when rendering realistic human bodies involving high-frequency details such as foliage, skin textures, hair, hands, and facial features.
Images generated using the Firefly Image 2 model have higher resolution, as well as more vivid colors and color contrast.
The Image 2 model also introduces new AI editing capabilities to help users customize results. Photo settings can manually or automatically adjust the depth of field, motion blur, and field of view of the resulting image. A "prompt guidance" function has also been added to help users improve the wording of text descriptions, while automatically completing prompts to improve efficiency.
Character.AI launches GroupChat, users can invite AI robots to join group chats
Character.AI, an AI robot startup founded by former Google AI researchers, recently launched the “Character Group Chat” function.
The feature provides users with a unique platform where they can not only chat with human companions but also add AI characters to the conversation.
These AI characters have unique personalities, knowledge, and abilities. Users can engage in group chats with AI characters based on historical figures, mythological gods, or any other background.
For example, you can have conversations with brilliant minds like Einstein, or explore mythical realms with gods like Zeus.
For more practical use cases, you can start group chats with friends around topics such as travel, gaming, or reading, and then invite AI characters to help facilitate and enhance these conversations.
The feature is first available to c.ai+ subscribers to gain feedback and make improvements, and users pay $9.9 per month to join the subscription plan.
Baidu will launch the first AI native map
According to the Shanghai Securities News, Baidu Vice President Shang Guobin revealed that Baidu will soon launch a new AI native map, fully applying the Wenxin large model to reconstruct the product interaction of the map.
According to reports, Baidu Map V19, which is reconstructed based on the Wenxin large model, will launch a new "AI guide". It has multiple rounds of natural language interaction capabilities, which is closer to the "real person communication" experience.
This "AI guide" also has a personalized digital human image. In addition to the default image recommended by the system, users can generate their own exclusive digital human by uploading pictures.
In addition, Baidu Maps’ lane-level navigation has also seen breakthrough progress recently.
Relying on the self-developed Beidou high-precision positioning technology, intelligent lane recommendation algorithm and 3D precision model-level high-definition rendering technology, Baidu Maps has newly upgraded urban lane-level navigation 3.0, which can provide clearer and three-dimensional global lane planning and guidance.
It is reported that Baidu Maps urban lane-level navigation will be launched in more than 100 cities by the end of the year, and will achieve nationwide coverage in 2024.
Hunt for insights｜Prophet
OpenAI scientists: GPT-4 is about to pass the inflection point
According to X user "Xiaohu"'s sharing, OpenAI scientist Hyung Won Chung said in a speech at Seoul National University:
GPT-4 is about to pass the inflection point and achieve a significant jump in its performance.
He believes that we need to change our mindset, because large language models (LLM) have huge potential, and some of its capabilities will only appear when the number of parameters reaches a certain scale.
Even if the current LLM still performs poorly in some aspects, we cannot rashly conclude that "it doesn't work." More accurately, it should be "it's not good yet."
Once the model scale is expanded, many original conclusions may be overturned.
Hyung Won Chung said LLM has the potential to continuously learn from failed experiments to improve problem-solving capabilities.
Sam Altman: "Sheltered" OpenAI is destined to die
According to Qubit, OpenAI CEO Sam Altman said in his speech at the latest YC alumni sharing session: "Shelling" OpenAI is destined to die.
Altman gave an example. For example, there are companies that are solving small flaws in the GPT model, especially focusing on the interface.
Firstly, OpenAI is already solving most of the work; secondly, the competitive advantage of these companies that only compete with OpenAI in terms of cost and speed will not be long-lasting and sustainable. Businesses must offer truly unique value.
In addition, Sam Altman said that he is optimistic about the empowerment of AI in medicine and education. AI medical consultants and AI personalized one-on-one tutoring, both of which have huge social value.
For example, in the education scenario, OpenAI is preparing to establish the OpenAI Academy, which is expected to be launched by the end of this year. It may be a free online teaching system accessible to everyone, and teachers can interact with GPT-5 and receive feedback and guidance during the course.
Prior to this, OpenAI also demonstrated two educational scenarios in GPT-4 customer cases. GPT-4 is transformed into an AI learning assistant, which can serve as both a virtual tutor for students and a courseware assistant for teachers.
Godfather of AI warns again: AI may replace humans
British computer scientist Geoffrey Hinton, known as the "Godfather of AI," warns that the dangers of AI may become more realistic than fiction.
Hinton left Google in May after a decade at Google, citing concerns about the risks of artificial intelligence and the potential for bad actors to use the technology.
Hinton said artificial intelligence can be both good and bad, and it’s unclear what will happen next. He believes that understanding the way technology thinks is as difficult as reading the human mind.
In the interview, he again warned that AI could replace humans. I'm not saying this will definitely happen. It would be great if we could stop them from doing this. But it's unclear whether we can stop them from wanting to do so.
Hinton believes that AI will help increase productivity and efficiency, but he is also concerned that many people may lose their jobs due to AI, and that there may not be enough jobs to replace those lost.
He said now is the time to conduct experiments to better understand artificial intelligence and pass laws to ensure the technology is used appropriately.
Hunt for Fun｜First
Image generation model competition
As mentioned before, Adobe released a number of image generation models this week. At the same time, DALL·E 3 was also officially launched some time ago. So which one is the level of the three image generation models: Midjourney, DALL·E 3 and Adobe Firefly? Which one is better?
For this problem, you might as well use the same Prompt (prompt word) to test. (From left to right are Midjourney, DALL·E 3 and Adobe Firefly)
Test X blogger said that Adobe Firefly did not even make any settings, but the generated effects were more diverse and realistic. So which image generation model do you think is better? Please share more opinions in the comment area.
Easter egg time
Which pair of shoes do you prefer
Author: Ethan Mollick
Tool: ChatGPT+DALL·E 3
Prompt word: GPT-4, show me product photoshoots for sneakers made for various famous people through history.
# Welcome to follow the official WeChat public account of aifaner: aifaner (WeChat ID: ifanr). More exciting content will be provided to you as soon as possible.