A day doesn't go by without a powerful open-source AI model drop.

 

DeepSeek (67B) makes up for the open-source shortcomings - math and coding!!

 

It beats Claude-2, scoring 65 on a National High School Exam.

 

DeepSeek is from China and is proof that the Chinese don't need our LLM tech; they can develop their own and are enlightened enough to open-source it!!

 

We are taking a look this week and will make it available in the fleszar-blog platform. 

Open source language models (LLMs) have the potential to catch up to models like GPT-4V and Google's Gemini, but it requires a collaborative effort from the open source community. Here are a few ways open source LLMs can catch up:

1. Community contributions: Open source LLMs benefit from contributions from a wide range of developers and researchers. Encouraging community involvement can lead to faster model improvements and innovations.

2. Data collection and curation: High-quality datasets are crucial for training LLMs. By organizing efforts to collect and curate relevant data, open source projects can improve the performance and accuracy of their models.

3. Optimization techniques: Researchers can develop novel optimization techniques specifically tailored for open source LLMs. These techniques can help improve model training efficiency and reduce computational requirements.

4. Transfer learning and pre-training: Leveraging pre-training techniques like transfer learning can help open source LLMs benefit from large-scale models like GPT-4V and Gemini. By fine-tuning these models on specific tasks, open source LLMs can reach higher levels of performance.

5. Collaboration with industry: Collaborating with industry experts and researchers can provide valuable insights and resources for open source LLM development. This collaboration can help bridge the gap between open source projects and advanced proprietary models.

It's important to note that catching up to models like GPT-4V and Gemini might require considerable time and resources. However, with continuous effort, community collaboration, and innovation, open source LLMs have the potential to narrow the gap and offer competitive alternatives.

Gemini is Google’s attempt to build a general-purpose AI program that can rival OpenAI’s GPT-4 model, which powers a paid version of ChatGPT. Demis Hassabis, the Google executive overseeing the project, told employees during a recent companywide meeting that the program would become available later this year, according to people who heard the remarks.

#artificialinteligence

 

Large Language Models (LLM ) are nothing more than machine learning models that are capable of performing a variety of Natural Language Processing (NLP) tasks. They are trained on huge data sets so that they are able to answer questions, generate their own content, properly classify it, summarize it or translate it into foreign languages.🍹

The appearance of their next generations of such models is proof of the rapid progress in the development of artificial intelligence.

It is estimated that the size of large language models has increased tenfold each year in recent years. As their size and, consequently, the level of complexity increase, so do their capabilities.

This is perfectly visible on the example of ChatGPT, which in its previous version was not so precise.🎺

--------------------------&---------------------------------

It couldn’t handle even longer written forms, it was often repetitive and as a result it didn’t deliver the values ​​expected by the end user. These imperfections have been largely eliminated in the currently available version of the model, but it is still far from perfect. However, this does not change the fact that its capabilities are impressive. Also impressive is the work done by the algorithms thanks to which it is able to surprise users and change our reality.

To build a model to support ChatGPT, OpenAI used a Microsoft-provided tens of millions of dollars worth of supercomputer , which at the time was among the top five most powerful machines in the world.

Thanks in advance 😭

Some things to worry about: In a bid to respond to your answers, it is very easy for the AI to “hallucinate” and generate plausible facts. It can generate entirely false content that is utterly convincing. Let me emphasize that: AI lies continuously and well. Every fact or piece of information it tells you may be incorrect. You will need to check it all. Particularly dangerous is asking it for math, references, quotes, citations, and information for the internet (for the models that are not connected to the internet). Bing and ChatGPT-4 are better at this.🎺

The AI also doesn’t explain itself, it only makes you think it does. If you ask it to explain why it wrote something, it will give you a plausible answer that is completely made up. It is not interrogating its own actions, it is just generating text that sounds like it is doing so. This makes understanding biases in the system very challenging, even though those biases almost certainly exist.💢

It also can be used unethically to manipulate or cheat. You are responsible for the output of these tools.

Large Language Models like ChatGPT are extremely powerful, but are built in a way that encourages people to use them in the wrong way.

The first thing people try to do with AI is what it is worst at; using it like Google: tell me about my company, look up my name, and so on. These answers are terrible. Many of the models are not connected to the internet, and even the ones that are make up facts. AI is not Google. So people leave disappointed.💐

Second, they may try something speculative, using it like Alexa, and asking a question, often about the AI itself. Will AI take my job? What do you like to eat? These answers are also terrible. With one exception, most of the AI systems have no personality, are not programmed to be fun like Alexa, and are not an oracle for the future. So people leave disappointed.

If people still stick around, they start to ask more interesting questions, either for fun or based on half-remembered college essay prompts: Write an article on why ducks are the best bird. Why is Catcher in the Rye a good novel? These are better.

As a result, people see blocks of text on a topic they don’t care about very much, and it is fine. Or the see text on something they are an expert in, and notice gaps. But it not that useful, or incredibly well-written. They usually quit around now, convinced that everyone is going to use this to cheat at school, but not much else.

---------------------------&----------------------------------

Al of these uses are not what AI is actually good at, and how it can be helpful. They can blind you to the real power of these tools. I want to try to show you some of why AI is powerful, in ways both exciting and anxiety-producing.🙋

Thanks in advance.

 

 

 

In a bold stroke, the EU’s amended AI Act would ban American companies such as OpenAI, Amazon, Google, and IBM from providing API access to generative AI models. The amended act, voted out of committee on Thursday, would sanction American open-source developers and software distributors, such as GitHub, if unlicensed generative models became available in Europe. While the act includes open source exceptions for traditional machine learning models, it expressly forbids safe-harbor provisions for open source generative systems.💐