Last week, meta drew attention by announcing that it could release a big language model (LLM) that may very well be used commercially. In February, Meta also announced ‘LLaMA’, a language model with relatively small parameters, as an open source, and since then, small language models using this as a foundational model have been pouring out.

Llama is available in 4 versions with 7 billion, 13 billion, 33 billion, and 65 billion parameters to pick from. Specifically, unlike large models comparable to OpenAI’s ‘GPT’ series or Google’s ‘Palm’, variables are relatively small, so supercomputing skills will not be required to operate.

Accordingly, first, researchers at Stanford University created the language model ‘Alpaca 7B’ in 3 hours using only $600 and 52,000 training data based on 7 billion ‘Lama’ variables.

Subsequently, small language models comparable to ‘Dolly’ and ‘Vicuna’ were developed one after one other, and a few developers created language models that may be run on laptops and mobile phones.

For that reason, the fundamental model of meta, lambda, is evaluated as leading the popularization of the language model. But even should you develop an application with this model, you’ll be able to’t run a business that makes money, since the meta side doesn’t allow lambdas for use for industrial purposes.

Because of this, a request has come from the industry to make it available for industrial use in addition to being released without spending a dime. An enthralling meme with the phrase ‘Free the LLaMa’ spread on the Web in an image of camelid llamas grazing in a meadow.

Meta initially rejected this request, saying, “We’re not ready yet.” It was a position that it was burdensome since the meta side could also take legal and moral responsibility for unintended effects if the open source model was used for industrial purposes.

Then, Meta said that he would allow the subsequent open-source language model for use commercially. On this case, developers will try to make use of the model that meta will provide you with, and the applying is prone to turn out to be a bot.

From a meta standpoint, a self-centered ecosystem may be built and technology development will also be helped. Technology evolves as more people use it, so you’ll be able to benefit from free feedback and enhancements. This process can also be the ability of open source.

While competitors comparable to OpenAI, Google, and Microsoft (MS) are developing large technologies or models and are concerned about copying, they have a tendency to turn out to be increasingly more secretive. It’s interpreted as

Next, we’ll let you know in regards to the technology trends.

■ Microsoft has announced that it’ll commercialize a quantum computer inside the subsequent 10 years. The form of particle that is vital to creating qubits, the fundamental unit of quantum, is ‘Majorana Fermion’, and it is alleged that they’ve developed a technology that may create and control it.

In the long run, Microsoft says it goals to construct quantum supercomputers to accurately design latest chemicals and materials.

■ Naver has set the name of the AI ​​chatbot, which is scheduled to be released around August after going through beta service next month, as ‘CUE:’. ‘Q:’ has applied for trademark registration with the Korean Mental Property Office as a service much like OpenAI’s ChatGPT.

The fundamental model of ‘Q:’ is named ‘OCEAN’, which is a model that’s specialized for search and adjusted ‘Hypaclova’, Naver’s large language model. In comparison with foreign-made AI chatbots, the quantity of learning Korean is 6,500 times greater.

■ Google released ‘Project Tailwind’, which routinely creates notes from memos or documents stored on the user’s computer. It is predicted to be useful for college kids taking study notes.

Because the user browses the document, Tailwind analyzes the text and displays details about related topics. Users may use easy prompts to ask Tailwind questions on document content and receive answers.

■ Global fintech company Tipalti investigated the recognition of 45 AI tools, and the result got here out that ‘ChatGPT’ got here in first place. It was ranked by items comparable to Google search volume, variety of SNS followers, and expert evaluation of software. The second place was ‘Copy.ai’, a tool for copywriting corporations.

As well as, tools developed within the US swept as much as twentieth place, with open source machine learning frameworks ‘PyTorch’ and ‘TensorFlow’ occupying third and 4th place, respectively.

Next, we’ll deliver key industry news.

■ Rutten Technologies, specializing in generative AI, attracted 15 billion won in Series A investment, bringing the cumulative investment to 19 billion won. This startup recently launched an app that auto-generates text and pictures for work, or a ‘tool builder’ that develops AI tools without coding.

Lutten has a history of becoming a hot topic by offering and choosing a ‘prompt engineer’ for the primary time in Korea with an annual salary of 100 million won. He revealed that he’ll recruit talent more aggressively with this investment.

■ Meta announced ‘Voicebox’, a generative AI model that creates audio from text. You may also perform speech synthesis, noise removal, editing and elegance conversion in 6 languages ​​and more.

Unlike existing models, it has been evaluated that it has taken a step forward in the sector of voice generation AI by developing a model that may perform various voice generation tasks without special training.

■ Mercedes-Benz has added ChatGPT function to its vehicles. With ChatGPT, he explained, drivers can ask for details about their destination, ask for a latest recipe for dinner, and even get answers to complex questions.

Nevertheless, foreign media identified that it’s questionable whether such a function is basically needed by the motive force. Some say that the service is geared toward the boss within the back seat slightly than the motive force.

■ Tesla CEO Elon Musk announced that he plans to start out a clinical trial to implant a pc chip into the human brain throughout the yr at Neuralink, a brain implant developer. US approval for the test was obtained last month.

A brain implant is an implantation of an electronic chip within the brain to enable communication with a pc, and is predicted to be a latest hope for the treatment of brain and nerve diseases comparable to blindness, paralysis, and dementia.

Reporter Jeong Byeong-il jbi@aitimes.com