- BrainBuzz
- Posts
- Apple introduces eight sleek AI language models.
Apple introduces eight sleek AI language models.
Aiming for on-device applications
Saidul Islam
April 26, 2024
Sponsored by
Welcome to BrainBuzz.
Today you will learn about Apple Eight Sleek AI-Language models are here, FREE AI Courses, Deepfake Camera is here, AI Tutorial, Alphabet providing dividend
And more…
AI Courses
Learn AI for FREE
Work lesser & drive 10x more impact using AI
HIGHLY RECOMMENDED: A Power-packed workshop (worth $199) for FREE and learn 20+ AI tools to become 10x more efficient at your work.
👉 Become an AI Genius in 3 hours. Register here (FREE for First 100) 🎁
In this workshop you will learn how to:
✅ Simplify your work and life using AI
✅ Do research & analyze data in seconds using AI tools
✅ Automate repetitive tasks & save 10+ hours every week
✅ Build stunning presentations & create content at lightning speed
AI News
Apple introduces eight sleek AI language models.
Apple has released a set of eight small AI language models, collectively named OpenELM (Open-source Efficient Language Models), aimed at on-device use. These models are small enough to run directly on a smartphone.
OpenELM’s Features and Specifications
The OpenELM models come in two types: four as “pretrained” and four as instruction-tuned. The models were trained on publicly available datasets and feature a 2048-token maximum context window. The models range from 270 million to 3 billion parameters, much smaller compared to other models like Meta’s Llama 3 family and OpenAI’s GPT-3.
OpenELM Models
The eight OpenELM models are:
OpenELM-270M
OpenELM-450M
OpenELM-1_1B
OpenELM-3B
OpenELM-270M-Instruct
OpenELM-450M-Instruct
OpenELM-1_1B-Instruct
OpenELM-3B-Instruct
OpenELM’s Potential Applications
While these models are mostly proof-of-concept research models for now, they could form the basis of future on-device AI offerings from Apple. The instruction-tuned models are more ideal for developing AI assistants and chatbots.
OpenELM’s Availability
The source code for OpenELM is currently available on the Hugging Face under an Apple Sample Code License. However, there are some restrictions in the license, so it may not fit the commonly accepted definition of "open source".
OpenELM’s Comparison with Other Models
This release mirrors efforts by Microsoft to make useful small AI language models that run locally. For instance, Microsoft’s Phi-3 models aim to achieve a similar level of language understanding and processing performance.
AI Deepfake
AI camera generates deepfakes by removing clothes.
The AI camera, named NUCA, developed by two German artists, can instantly remove a person’s shirt or any other clothing in real time. This camera uses a smartphone to capture images which are then processed by AI algorithms in the cloud.
How NUCA Works
NUCA reconstructs what it predicts a person’s naked body would look like based on data about their gender, face, age, and body shape. The process is quick and requires minimal technical knowledge, completing its process in about 10 seconds.
The Dark Side of NUCA
While NUCA could be seen as a magic editing tool for creatives, it also highlights the potential misuse of AI technology to infringe on personal privacy. The ease and speed with which NUCA operates amplify the potential for misuse.
Ethical Concerns Surrounding NUCA
Without requiring consent to digitally “undress” someone, NUCA could easily be used for malicious purposes such as blackmail or cyberbullying. The artists hope that by exposing this capability, they will ignite a public debate on the ethical trajectory of AI development.
NUCA’s Availability and Potential Misuse
While the artists do not intend for NUCA to be used commercially, the technology it represents could be replicated and used unethically by others. This raises profound implications for social norms and personal security as AI capabilities grow.
AI Tutorial
7 Reasons Why Copilot Outshines ChatGPT
1. No Knowledge Cutoff
The most significant advantage of Copilot over ChatGPT is its lack of a knowledge cutoff. Copilot, powered by Microsoft’s Bing search engine, has access to the most up-to-date information on the internet. In contrast, the free version of ChatGPT only has access to the information it was trained on, limiting its knowledge base to information up to January 2022.
2. Footnotes and Cited Sources
Copilot includes footnotes with every response that cites the website the chatbot pulled its answer from. This feature makes it easier for users to further their research and confirm the accuracy of the chatbot’s response.
3. Advanced Model
The free version of Copilot uses a more advanced version of OpenAI’s large language model compared to the GPT-3.5 used by the free version of ChatGPT.
4. Free Access to Premium Features
All the features discussed above are available on the free version of Copilot. Many of these features are limited to the premium version of ChatGPT, ChatGPT Plus, which costs users $20 a month.
5. Real-Time Information
Copilot can provide real-time information. For example, when asked about the result of a recent sports game, Copilot was able to provide an accurate answer, while ChatGPT could not.
6. Accuracy of Information
Even for events before the knowledge cutoff, there is no guarantee that ChatGPT will have the most accurate response as it doesn’t know about recent events that could have impacted the situation.
7. Recognition and Credit
Despite its advantages, Copilot doesn’t get the credit it deserves. The author of the article believes that more users might switch to Copilot if they were aware of these benefits.
AI STOCKS
Alphabet stock soars on earnings beat, dividend news.
Alphabet, Google’s parent company, reported first quarter earnings that exceeded expectations. The company reported an adjusted earnings per share of $1.89, beating consensus estimates of $1.53. Revenue, excluding traffic acquisition costs, rose 16% from the same period last year to $67.59 billion, surpassing analyst expectations of $66.07 billion.
Dividend Announcement and Stock Buyback
In addition to the earnings beat, Alphabet announced a cash dividend program of $0.20 per share. The board of directors also approved stock repurchases of up to an additional $70 billion.
AI Development and Monetization
CEO Sundar Pichai emphasized Alphabet’s commitment to AI development. He stated that the company has clear paths to monetize AI breakthroughs through advertising, cloud, and subscriptions. He also highlighted the integration of AI tools into Google search, which allows users to ask more complex and descriptive questions.
Cloud Revenue and Ad Revenue Growth
For the second straight quarter, cloud revenue rose nearly 30% from the year-earlier period to surpass $9 billion. Ad revenue, the heart of Google’s business, rose by 13%, registering $61.66 billion.
Capital Expenditures and AI Investments
Alphabet reported capital expenditures for the quarter of $12 billion, mostly tied to servers and data centers. CFO Ruth Porat said during the call that the quarters ahead will register similar levels of spending, reflecting the company’s confidence in its AI investments.
Breaking News
Meta’s stock value dropped by $200B as CEO Zuckerberg’s focus on long-term AI and metaverse investments unnerved investors.
Los Angeles Metro will use AI-powered cameras to identify and ticket vehicles parked in bus lanes, with testing starting this summer.
Saudi Arabia is investing heavily in AI, aiming to become a global superpower in the field, amidst escalating worldwide competition.