Meta enters the world of AI chatbots with LLaMA
Like many other large tech companies, Meta is working on developing its own machine learning model that can be used for various applications. At the end of February, LlaMA was presented, a model that is based on 65 billion parameters and makes decisions. Although the model was only made available to selected test groups, it is already being shared on the internet via torrent and there is a Github repository with installation guides for various operating systems.
The first tests by the community show that LlaMA works on the same level as Text-davinci, which in turn is based on OpenAI’s GPT-3 model. So, Metas Software can generate relatively accurate pieces of text based on user input.
Meta enters the world of AI chatbots with LLaMA
Like many other large tech companies, Meta is working on developing its own machine learning model that can be used for various applications. At the end of February, LlaMA was presented, a model that is based on 65 billion parameters and makes decisions. Although the model was only made available to selected test groups, it is already being shared on the internet via torrent and there is a Github repository with installation guides for various operating systems.
The first tests by the community show that LlaMA works on the same level as Text-davinci, which in turn is based on OpenAI’s GPT-3 model. So, Metas Software can generate relatively accurate pieces of text based on user input.
Meta focuses on artificial intelligence
Meta focuses on artificial intelligence
At Meta, the current focus is on artificial intelligence and the development of a chatbot, while the Metaverse has taken a back seat. Mark Zuckerberg has now announced that internal teams are being restructured to work more efficiently. A top-level group will be created at Meta to focus on Generative AI to make faster progress in this area. People from different departments who already have experience with AI should be brought together to integrate this technology into the different products and create great experiences.
Initially, the focus will be on the development of creative tools, but in the long term, AI models will also be developed that can help in a variety of ways, for example in texts for WhatsApp and Messenger, images and filters, and videos for Instagram. However, there is still a lot of work to be done before these forward-looking experiences are available to users. Meta had made similar attempts in the past but failed. After the last focus was on the Metaverse, the company is now apparently returning to artificial intelligence.
At Meta, the current focus is on artificial intelligence and the development of a chatbot, while the Metaverse has taken a back seat. Mark Zuckerberg has now announced that internal teams are being restructured to work more efficiently. A top-level group will be created at Meta to focus on Generative AI to make faster progress in this area. People from different departments who already have experience with AI should be brought together to integrate this technology into the different products and create great experiences.
Initially, the focus will be on the development of creative tools, but in the long term, AI models will also be developed that can help in a variety of ways, for example in texts for WhatsApp and Messenger, images and filters, and videos for Instagram. However, there is still a lot of work to be done before these forward-looking experiences are available to users. Meta had made similar attempts in the past but failed. After the last focus was on the Metaverse, the company is now apparently returning to artificial intelligence.
Meta’s bulky abbreviation for AI language model “LLaMA”
Meta’s bulky abbreviation for AI language model “LLaMA”
Meta recently presented LLaMA (Large Language Model Meta AI), a language model intended to serve as the basis for its own chatbot. It is more powerful than OpenAI’s GPT-3 and Google’s PaLM in terms of the number of parameters used during training. LLaMA comes in four versions with different sets of parameters, although developers say even the smallest model can provide better answers than the competition. However, LLaMA is currently only available to researchers on a waitlist. The release of a chatbot for the public seems to be a long way off.
A few years ago, Meta, then known as Facebook, tried to integrate chatbots into its networks for various functions. The plan was that you could, for example, buy cinema tickets directly in Messenger by asking the chatbot to do so by entering text, which then automatically books and pays. Unfortunately, this did not work and the function was never implemented. Although there are shopping options across the various platforms, customers still have to manually navigate through the Services to complete a purchase. Meta’s previous chatbot, the BlenderBot, which is open at least for testing purposes, shouldn’t repeat previous problems of AI-based bots, such as bias and inappropriate responses. However, this has only had limited success so far.
It remains unclear whether the merger of employees will also result in job cuts. Zuckerberg recently said last year’s job cuts were just the beginning. In this year of efficiency, middle management is set to be trimmed and layoffs are expected to occur in nearly every area. Meta needs to get back on its feet financially and keep investors engaged. In addition to misjudgments of the business in the booming Corona times, inflation and collapsing advertising revenues are causing the company to plummet.
Meta recently presented LLaMA (Large Language Model Meta AI), a language model intended to serve as the basis for its own chatbot. It is more powerful than OpenAI’s GPT-3 and Google’s PaLM in terms of the number of parameters used during training. LLaMA comes in four versions with different sets of parameters, although developers say even the smallest model can provide better answers than the competition. However, LLaMA is currently only available to researchers on a waitlist. The release of a chatbot for the public seems to be a long way off.
A few years ago, Meta, then known as Facebook, tried to integrate chatbots into its networks for various functions. The plan was that you could, for example, buy cinema tickets directly in Messenger by asking the chatbot to do so by entering text, which then automatically books and pays. Unfortunately, this did not work and the function was never implemented. Although there are shopping options across the various platforms, customers still have to manually navigate through the Services to complete a purchase. Meta’s previous chatbot, the BlenderBot, which is open at least for testing purposes, shouldn’t repeat previous problems of AI-based bots, such as bias and inappropriate responses. However, this has only had limited success so far.
It remains unclear whether the merger of employees will also result in job cuts. Zuckerberg recently said last year’s job cuts were just the beginning. In this year of efficiency, middle management is set to be trimmed and layoffs are expected to occur in nearly every area. Meta needs to get back on its feet financially and keep investors engaged. In addition to misjudgments of the business in the booming Corona times, inflation and collapsing advertising revenues are causing the company to plummet.
Popular Posts:
Ad-free home network: Install Pi-hole on Windows
Say goodbye to ads on smart TVs and in apps: Pi-hole software turns your Windows laptop into a network filter. This article explains step-by-step how to install it via Docker and configure the necessary DNS settings in your FRITZ!Box.
How to tune your FRITZ!Box into a professional call server
A professional telephone system can be built using a FRITZ!Box and a laptop. This article shows step by step how to use the free software "Phoner" to schedule announcements and record calls – including important legal information (§ 201 StGB).
Why to-do lists are a waste of time
Do you feel unproductive at the end of the day, even though you've worked hard? Your to-do list is to blame. It tempts you to focus on easy tasks and ignores your limited time. This article explains why lists are "self-deception" and why professionals use a calendar instead.
Smartphone Wi-Fi security: Public hotspots vs. home network
Is smartphone Wi-Fi a security risk? This article analyzes in detail threats such as evil twin attacks and explains protective measures for when you're on the go. We also clarify why home Wi-Fi is usually secure and how you can effectively separate your smart home from sensitive data using a guest network.
Warum dein Excel-Kurs Zeitverschwendung ist – was du wirklich lernen solltest!
Hand aufs Herz: Wann hast du zuletzt eine komplexe Excel-Formel ohne Googeln getippt? Eben. KI schreibt heute den Code für dich. Erfahre, warum klassische Excel-Trainings veraltet sind und welche 3 modernen Skills deinen Marktwert im Büro jetzt massiv steigern.
Cybersicherheit: Die 3 größten Fehler, die 90% aller Mitarbeiter machen
Hacker brauchen keine Codes, sie brauchen nur einen unaufmerksamen Mitarbeiter. Von Passwort-Recycling bis zum gefährlichen Klick: Wir zeigen die drei häufigsten Fehler im Büroalltag und geben praktische Tipps, wie Sie zur menschlichen Firewall werden.
Offers 2024: Word & Excel Templates
Popular Posts:
Ad-free home network: Install Pi-hole on Windows
Say goodbye to ads on smart TVs and in apps: Pi-hole software turns your Windows laptop into a network filter. This article explains step-by-step how to install it via Docker and configure the necessary DNS settings in your FRITZ!Box.
How to tune your FRITZ!Box into a professional call server
A professional telephone system can be built using a FRITZ!Box and a laptop. This article shows step by step how to use the free software "Phoner" to schedule announcements and record calls – including important legal information (§ 201 StGB).
Why to-do lists are a waste of time
Do you feel unproductive at the end of the day, even though you've worked hard? Your to-do list is to blame. It tempts you to focus on easy tasks and ignores your limited time. This article explains why lists are "self-deception" and why professionals use a calendar instead.
Smartphone Wi-Fi security: Public hotspots vs. home network
Is smartphone Wi-Fi a security risk? This article analyzes in detail threats such as evil twin attacks and explains protective measures for when you're on the go. We also clarify why home Wi-Fi is usually secure and how you can effectively separate your smart home from sensitive data using a guest network.
Warum dein Excel-Kurs Zeitverschwendung ist – was du wirklich lernen solltest!
Hand aufs Herz: Wann hast du zuletzt eine komplexe Excel-Formel ohne Googeln getippt? Eben. KI schreibt heute den Code für dich. Erfahre, warum klassische Excel-Trainings veraltet sind und welche 3 modernen Skills deinen Marktwert im Büro jetzt massiv steigern.
Cybersicherheit: Die 3 größten Fehler, die 90% aller Mitarbeiter machen
Hacker brauchen keine Codes, sie brauchen nur einen unaufmerksamen Mitarbeiter. Von Passwort-Recycling bis zum gefährlichen Klick: Wir zeigen die drei häufigsten Fehler im Büroalltag und geben praktische Tipps, wie Sie zur menschlichen Firewall werden.
Offers 2024: Word & Excel Templates
