In partnership with

AI World Update for July 29 - August 4, 2024

I haven't written a news update for a long time because there is a lot of work on weekdays. I couldn't do it in time, so I used my vacation to sit and review. Various updates

Last week, the AI world was called an update that could barely keep up.

A start.

1. OpenAI Upgrades GPT-4o Model to Support Longer Context Up to 64,000 Tokens

• Now we can throw something long, a larger file to it. From the original, you can get 4,000 tokens, now it has expanded to 64,000 tokens, or 16 times.
• What can be done with a larger context? The conclusions from stuffing in big information will be much better. The seminars were long all day and then transcribed into a script. It will be easier to send it to summarize it, no need to cut it into several pieces. What about doing research that requires too many research papers for humans to finish in a short time, or creating a better quality article? From stuffing in the information and sending it to him to study, he will also write better.

1. OpenAI ต่อ กับ Advanced Voice Mode

• OpenAI began rolling out the Advanced Voice Mode feature that was shown to the world a few months ago. I can talk to people very naturally. But we only saw it in the clip, not actually using it. This time around. It is now open to others only for some ChatGPT Plus users (it is called Alpha 55).
• This mode It's called Hyper Realistic Audio, which can stop waiting for the rhythm of a conversation naturally...

1. Google ปล่อยโมเดล Gemini 1.5 Pro

• After the release of Gemini 1.0 Pro in February, Google was silent. I'm a language model, called opening up to big form, but actually playing on the field. (Actually, if you play behind the scenes, Google AI Studio, it will turn off these sensors).
• This time, Gemini 1.5 Pro (Experimental 0801) is back to reclaim the No. 1 throne of Benchmark (tested by LMSYS, which regularly updates the leaderboard of each LLMs). Some people don't believe the test from this camp.

• The test is probably a test, a specific test with some tasks, but when it is actually used, there are people who scold 555 ...
• Gemini 1.5 Pro is a smarter, multimodal model that supports text, images, audio, and video as before, while expanding the Windows context. Increased to 2,000,000 tokens, 2 times larger than Gemini 1.0 Pro.
• Complex tasks for businesses will also be much better because they can process more data. Especially complex data analysis work. Software Development

4. Google releases Gemma 2 model to run on devices

• The big brother model, Gemini 1.5 Pro, has been released, and the younger model, Gemma 2, which can run on various devices (called Gemma 2 2B, is trained with a dataset of 2 billion parameters, compared to GPT3.5, which uses 175 billion).
• It is called a model that is sufficient for general use. It uses dozens of times less resources, but the ability is so comparable that it can be stuffed into various devices.
• Gemma 2 opens in 2 types, a very small one (Gemma 2 2B) and a slightly larger but still small one (Gemma 2 27B).

1. Let's take a look at Meta with the new Segment Anything Model or SAM 2.

• Previous Week Brother Mark played big, releasing the Llama 3.1 70B and 405B to shake the industry.
• This week, I still can't stop with the very popular model. Segment Anything Model or SAM 2 with the ability to distinguish objects in photos and videos.
• For iPhone and iPad users, think about when we open a photo, touch a person, a plate, or any object in the photo with our finger, and drag and drop it into other apps. That kind of emotion.
• But SAM2 it can already do with video. It is said that it will bring a lot of benefits to the video industry because it can select objects in the video or prompt and change them in real time. No need to wait for a new video render.
• You can try it here. https://sam2.metademolab.com/demo  

1. Midjourney released new version 6.1

• The quality of the photos it produces is much more beautiful than before, especially the portraits of people who look like they are indistinguishable. Arms, legs, body, people, skin textures.
• Jane 25% faster

1. Runway ออกโมเดล Image-to-Video รุ่น 3

• Faster, cheaper than previous models.
• There are 2 models, the Gen 3 Alpha and the Gen 3 Alpha Turbo, which are 7 times faster than the regular model.

• The Generative Video Battleground is super hot. Let's play the debut and collide. Pika Labs, Dream Machine from Luma AI, Kling from China, and Sora from OpenAI.
• From the original when Sora came out new. People were so excited that Lumier from Google disappeared.

Now, everyone has developed themselves to the point where their abilities are close to each other, and the service fee is cute and wallet-friendly (I thought it would come for $50 per month), and I found Kling of China for only $5 per month. https://klingai.com/

1. Canva Acquisitions Leonardo.ai

• People in the Creator industry are quite excited because: Leonardo.ai This is a popular image and video generation tool. It is not very difficult to use. Create high-quality images and videos virtually. It is used by about 19 million people (55 is a lot).
• It is a good purchase to strengthen the core product, which Canva has been buying related businesses continuously. This year, I just bought it. Affinity that makes programs similar to Adobe Photoshop and Illustrator are added.
• We signed up for Canva cheaply and used it very cost-effectively. It's getting better and better, and we're still paying the same (let's stay like this for a long time).

1. Paris 2024 Olympics with AI

• People may not know much about it. In this Olympics, there are quite a lot of applications of AI.
• From video editing to highlighting to automatic AI cutting, and also resizing to multiple formats to various online platforms, this platform is owned by Intel called Intel Geti AI.
• It also uses AI to dub and translate into other languages. In order to really reach the whole world.
• AI is used to monitor accounts on social media. Hundreds of thousands of accounts to check the message, accuracy, or fake news.
• The AI Chatbot has been tested on the Athlete365 platform ( https://olympics.com/athlete365 ) also to help answer questions about various rules of record in competitions, drug use monitoring rules.
• Digital Twin technology is also used to simulate various competitions.
• Camera systems used in competitions It has partnered with Alibaba to record clips and replays, as well as use AI to create 3D images, simulate sports competitions, and process multiple camera angles to see all angles.
• The size of the Omega watch used in the Olympics for timing also adds an AI function that can be used in conjunction with the camera to collect data such as running to the finish line, swimming, etc. High jump, cycling and then using it as data to analyze and compare performance.
• This Olympics is the first AI-powered Olympics to use AI in almost every part of the ! ....event. The stock fell sharply and recently laid off 15,000 employees.

1. Finally, it is a matter of Responsible AI between Brazil and Meta companies.

• Brazil's National Data Protection Authority (ANPD) is responsible for security and data privacy in Brazil. Meta has banned the use of AI in Brazil after ordering the company to The personal data of more than 102 million Brazilians who use Facebook and Instagram is prohibited from being used for AI training.
• But Meta still doesn't strictly follow the rules. He refused to clarify what information he took and what he did, until he was banned.
• I want our house to be more intensive with these platforms. Because we don't know what information they use to train their AI.

สรุปงาน SCBX Unlocking AI EP 10: Responsible AI in Action: From Regulation to Real-World Impact

1. The AI Regulatory Landscape: Navigating Current and Future Framework  
   by Dr. Sak Sekkuntad Digital Transformation Consultant from the Electronic Transaction Development Agency (EDA)
   
2. AI Success Story: Implementing Responsible AI Across Industries Around the Globe
   by Ms. Kritiyanee Burantrivet Baker McKenzie Attorneys & Partners
   
3. How to Maximize AI Transformation in Corporate from Day 1 to Scale Out
   by Mr. Nattapol Jongcharoonkiet Vice President of PTT Global Chemical Public Company Limited or GC
   
4. Demarginalizing the Underserved with AI Lending  
   by Mr. Thiranan Arunwattankul Chief Operating Officer, MONIX Co., Ltd.
   
5. Kickstarting All In Organizations: Creating Responsible AI Policies and Best Practices for Compliance

This newsletter is sponsored by Intercom.com Global Customer Support Software Company

They are giving out 90% discounts to startups around the world. I recommend you to hurry up and get a 90% discount because their Software as a Service system is used all over the world.

Intercom for Startups

Join Intercom’s Early Stage Program to receive a 90% discount.

Get a direct line to your customers. Try the only complete AI-first customer service solution.

Apply now

Download the article in presentation format here.