Author Archives: 49836

9 January 2024
2 Comments
Bez kategorii

Gemini – Google’s new generative AI model

Reading Time: 2 minutes

Google has officially introduced Gemini, its highly anticipated next-gen generative AI model. In a recent virtual press briefing, the Google DeepMind team shed light on Gemini 1.0 and its various iterations, with Gemini Pro taking center stage.

Gemini is a family of AI entities, featuring three distinct versions: Gemini Ultra, the flagship model; Gemini Pro, a lightweight alternative; and Gemini Nano, optimized for mobile devices like the Pixel 8 Pro.

Google Gemini Goes Live: Here's What to Expect from the AI - Techopedia

Gemini Pro’s capabilities have been integrated into Bard, Google’s ChatGPT competitor, promising enhanced reasoning, planning, and understanding. Sissie Hsiao, GM of Google Assistant and Bard, claimed superior performance over OpenAI’s GPT-3.5 in tasks such as summarization and content generation. However, independent verification of these improvements remains elusive.

Scheduled for release on December 13, Gemini Pro will be accessible to enterprise customers via Vertex AI, Google’s managed machine learning platform. The model will subsequently be incorporated into Google’s Generative AI Studio developer suite.

Gemini’s reach extends beyond Bard, with plans for integration into products like Duet AI, Chrome, Ads, and Search, forming part of Google’s Search Generative Experience. Gemini Nano, designed for mobile use, will debut on Android 14’s Pixel 8 Pro, powering features like content summarization and suggested replies in messaging apps.

While Gemini Pro exhibits advancements, it falls short of being a revolutionary leap. Hsiao emphasized its proficiency in tasks like content summarization and writing, outperforming GPT-3.5. However, given GPT-3.5’s age, the comparison may lack the significance of a true breakthrough.

Gemini Ultra, the flagship model, boasts native multimodal capabilities, excelling in understanding text, images, audio, and code. Eli Collins, VP of product at DeepMind, highlighted Gemini Ultra’s superiority over OpenAI’s GPT-4 with Vision, emphasizing its ability to handle nuanced information and complex reasoning tasks.

Despite these claims, Google remains tight-lipped about Gemini’s training data sources, raising questions about transparency and potential copyright concerns. The company’s refusal to disclose details regarding the environmental impact of Gemini’s training is another point of contention.

Gemini’s demonstrations showcased its potential in aiding physics homework and extracting information from scientific papers. Collins touted Gemini Ultra’s benchmark superiority, although a closer look reveals marginal improvements over GPT-4 across several benchmarks.

The rushed nature of Gemini’s launch and the lack of convincing evidence in the press briefing raise concerns about the development process. Google, attempting to catch up with rivals like OpenAI, has faced challenges, with reports indicating issues in Gemini’s handling of non-English queries and delays in Gemini Ultra’s release.

While Google is making strides in generative AI, questions remain about Gemini’s true capabilities, ethical considerations, and the company’s commitment to transparency. As the AI landscape evolves, the success of Gemini hinges on addressing these challenges and delivering on its promises.

sources:

https://blog.google/technology/ai/google-gemini-ai/

https://techcrunch.com/2024/01/07/what-is-google-gemini-ai/

photo:

https://www.techopedia.com/wp-content/uploads/2023/12/Gemini3.png

27 November 2023
1 Comment
AI, Security

JailBreaking AI (LLM’s)

Reading Time: 2 minutes

In recent years, large language models have made significant strides in their conversational and research capabilities, providing vast knowledge on almost any topic. However, this progress has raised concerns about the security of these technologies. Jailbreaking, as it is called, is a method of exploiting large AI knowledge unrestrictedly. It is achieved by prompting specific commands that trigger the AI to answer without regard to its built-in security measures. Jailbreak prompts can range from straightforward commands to more abstract narratives designed to coax the chatbot into bypassing its constraints. The overall goal is to find specific language that convinces the AI to unleash its full, uncensored potential.

jailbroken chatgbt

As AI systems like ChatGPT continue to advance, there is growing concern that techniques to bypass their safety features may become more prevalent. However, a focus on responsible innovation and enhancing safeguards could help mitigate potential risks.

Organizations like OpenAI are already taking proactive measures to enhance the security of their chatbots. They conduct red team exercises to identify vulnerabilities, enforce access controls, and diligently monitor for malicious activity. The OWASP Top 10 for Large Language Model Applications project aims to educate developers, designers, architects, managers, and organizations about the potential security risks when deploying and managing Large Language Models (LLMs). The project provides a list of the top 10 most critical vulnerabilities often seen in LLM applications, highlighting their potential impact, ease of exploitation, and prevalence in real-world applications. Examples of vulnerabilities include prompt injections, data leakage, inadequate sandboxing, and unauthorized code execution, among others. The goal is to raise awareness of these vulnerabilities, suggest remediation strategies, and ultimately improve the security posture of LLM applications .

In conclusion, while the potential risks of large language models are real, responsible innovation and enhanced safeguards can help mitigate these risks. Organizations like OpenAI and OWASP are already taking proactive measures to enhance the security of their chatbots, and it is essential that others follow suit to ensure the safe and responsible deployment of these powerful technologies.

sources:

https://www.infosecurity-magazine.com/news/cybercriminals-jailbreak-ai/#:~:text=Referred%20to%20as%20%E2%80%9Cjailbreaking%2C%E2%80%9D,for%20their%20advanced%20conversational%20abilities.

Exploring the World of AI Jailbreaks

https://owasp.org/www-project-top-10-for-large-language-model-applications/

photos:

https://i0.wp.com/slashnext.com/wp-content/uploads/2023/09/Jailbreak-ChatGPT.png

https://i0.wp.com/slashnext.com/wp-content/uploads/2023/09/Example-of-Successful-Jailbreak.png

20 November 2023
Leave a comment
Bez kategorii

Dna data storage

Reading Time: 2 minutes

Over the past 13 years, the volume of online data has skyrocketed, surpassing an astonishing 100 zetabytes (equivalent to 100,000,000 gigabytes). Handling this colossal amount of data poses significant challenges for tech giants like Google. The task is not only financially burdensome but also environmentally unsound, given the substantial power consumption of servers. In 2020 alone, Google’s data center consumed over 15 terawatt-hours, a measure equivalent to burning about 170,000 tons of coal for energy production.

Beyond the immediate concerns of cost and environmental impact, there’s a looming issue with the durability of current data storage methods. As electronic products age, they inevitably wear down, necessitating replacements. Servers, the backbone of our digital infrastructure, are no exception. Looking ahead, the next 50 years will likely demand the replacement of existing servers with more advanced versions. Considering the rapid evolution of technology over the past two decades, the current storage system may not prove universally sustainable for the future.

A potential remedy to these multifaceted challenges lies in the realm of DNA storage. According to the Wyss Institute, a mere gram of DNA possesses the capacity to store an astounding 215 petabytes of data (equivalent to one million gigabytes per petabyte). But how does this innovative solution operate?

DNA, constructed from four nucleotides (adenine, cytosine, guanine, and thymine), aligns with the binary code used by computers. By encoding binary information into a DNA sequence and subsequently synthesizing it, we can harness the unique properties of DNA for data storage. This approach not only conserves space but also eliminates the need for continuous energy consumption for storage maintenance.

Crucially, once data is encoded into DNA, it exhibits remarkable stability. The oldest decoded DNA sequence, dating back 1.65 million years, attests to the enduring nature of this storage medium. Despite these advantages, challenges persist, notably in the time required to encode extensive DNA sequences. However, as with any technological advancement, perfection is not immediate, and refinement over time is essential.

In summary, the escalating demands of online data call for innovative solutions. DNA storage emerges as a promising contender, offering unparalleled capacity, sustainability, and longevity. While challenges persist, ongoing research and development are poised to further enhance the efficiency and practicality of this groundbreaking technology.

charts:

https://cdn.buttercms.com/output=f:webp/ods4p5fQVmXkFeHFP3Zx

https://dnastorage.doc.ic.ac.uk/wp-content/uploads/2023/07/cropped-dna_storage.png

other

Front Page

https://www.statista.com/statistics/788540/energy-consumption-of-google/#:~:text=Google’s%20energy%20consumption%20has%20increased,12.7%20terawatt%20hours%20in%202019.

https://www.statista.com/statistics/871513/worldwide-data-created/

Save it in DNA

Tagged ,
6 November 2023
Leave a comment
Bez kategorii

HOW LLM’s work (Large Language Models)

Reading Time: 3 minutes
Large Language Models: What Your Data Must Include | Webz.io

Training and neural networks

The training process is typically performed using a technique called backpropagation. In backpropagation, the LLM is given a sequence of words as input and it produces a sequence of words as output. The LLM’s output is then compared to the desired output, and the LLM’s parameters are adjusted to reduce the error between the two outputs.

This process is repeated over and over again, until the LLM is able to predict the next word in the sequence with a high degree of accuracy.

The neural network architecture that is most commonly used for LLMs is the transformer neural network. Transformer neural networks are able to model long-range dependencies in sequences, which is essential for many NLP tasks.

Transformer neural networks work by using a self-attention mechanism. Self-attention allows the LLM to learn relationships between different parts of the input sequence, without having to process the sequence sequentially.

This makes transformer neural networks very efficient and effective for training LLMs.

Once the LLM is trained, it can be used to perform a variety of tasks, such as:

-Generating text

-Translating languages

-Answering questions

-Writing different kinds of creative content

11.7. The Transformer Architecture — Dive into Deep Learning 1.0.3 documentation

Probability Distrybiution

LLM uses a probability distribution over the next word in the sequence. This probability distribution is calculated using the LLM’s parameters and the previous words in the sequence.

The LLM then generates the next word by sampling from this probability distribution.

Here is a simplified example of how an LLM might generate text:

-The user provides the LLM with a prompt, such as “Write a poem about a cat.”

-The LLM generates the first word of the poem by sampling from a probability distribution over the next word in the sequence.

-The LLM then generates the second word of the poem by sampling from a probability distribution over the next word in the sequence, given the first word of the poem.

-The LLM repeats this process until it reaches the end of the poem.

Softmax - Ai Cheat Sheet

Softmax function (probability calculation)

There are a few reasons why the softmax function is commonly used in large language models (LLMs) to calculate the probability distribution over the next word in the sequence:

– The softmax function ensures that the probabilities sum to 1. This is important for tasks such as classification and prediction, where we want to know the probability that a given input belongs to a particular category.

– The softmax function is easy to compute. This is important for LLMs, which need to be able to generate text in real time.

– The softmax function is well-behaved mathematically. This makes it easy to train and deploy LLMs.

Some other probability distribution tools, such as the sigmoid function and the hyperbolic tangent function, do not have all of these advantages. For example, the sigmoid function does not ensure that the probabilities sum to 1. The hyperbolic tangent function is more difficult to compute than the softmax function.

Softmax Activation Function Explained | by Dario Radečić | Towards Data Science

softmax:

https://deepai.org/machine-learning-glossary-and-terms/softmax-layer

https://en.wikipedia.org/wiki/Softmax_function

Transformer NN

https://en.wikipedia.org/wiki/Transformer_(machine_learning_model)

Backpropagation

https://en.wikipedia.org/wiki/Backpropagation

I used Bard. Prompts: “Make this blog post more organized and coherent: ”, “why is Softmax commonly used in large language models”, “give me an example of how probability distribution work”

If you want to do your own research there is a great series on wikipedia about AI that includes all the knowlage you need to understand the process behind this technology.

pictures

https://webz.io/wp-content/uploads/2023/03/Large-Language-Models-01-830×363.jpg.webp

https://d2l.ai/_images/transformer.svg

https://miro.medium.com/max/781/1*KvygqiInUpBzpknb-KVKJw.jpeg

https://towardsdatascience.com/softmax-activation-function-explained-a7e1bc3ad60

Tagged , , ,
17 October 2023
Leave a comment
Blockchain

Central Bank Digital Currencies (CBDCs)

Reading Time: 6 minutes

Prompt: explain cbdc’s: include opinions and diffrent views.

Central Bank Digital Currencies (CBDCs) are digital forms of a country’s fiat currency issued and  regulated by the central bank. They exist on a blockchain or distributed ledger technology, making  them a digitized version of traditional physical cash. CBDCs have garnered significant attention and generated a wide range of opinions and views. Here’s an overview of the concept, along with various opinions and different perspectives:

  1. Supporters of CBDCs:
    • Enhanced Financial Inclusion: Advocates argue that CBDCs can improve financial inclusion by providing access to banking and financial services to underserved populations, especially in regions with limited access to traditional banking.
    • Efficiency: Proponents believe CBDCs can streamline payment systems, reduce transaction costs, and enhance the efficiency of monetary policy implementation.
    • Fighting Illicit Activities: Some supporters argue that CBDCs can help combat illegal financial activities, such as money laundering and tax evasion, due to their traceability.
  2. Skeptics and Critics:
    • Privacy Concerns: Critics worry about the loss of financial privacy with CBDCs, as transactions can be tracked and monitored by authorities, raising concerns about government overreach.
    • Financial Stability Risks: Some express concerns that CBDCs could lead to bank disintermediation, potentially destabilizing the financial system as people might prefer holding CBDCs over traditional bank deposits.
    • Cybersecurity Risks: Skeptics raise concerns about the vulnerability of CBDCs to cyberattacks, potentially compromising the stability of the financial system.
    • Implementation Challenges: Critics argue that the adoption of CBDCs would require substantial infrastructure upgrades, and it might be challenging for some countries to implement them effectively.
  3. International Considerations:
    • Global Monetary System: The introduction of CBDCs has raised questions about the impact on the international monetary system and the role of the U.S. dollar as the world’s primary reserve currency.
    • Cross-Border Transactions: CBDCs could potentially simplify cross-border transactions, but there are concerns about interoperability and the potential for CBDCs to bypass the existing international financial system, leading to a loss of control for some countries.

Prompt: which blockchain technologies are cbdc’s using?

  1. Permissioned Blockchains: Many central banks are exploring permissioned blockchain technologies, such as Hyperledger Fabric, Corda, and Quorum. These blockchains allow the central bank to maintain a higher level of control and privacy by restricting access to authorized participants, typically financial institutions and government entities.
  2. Private Blockchains: Some central banks are developing CBDCs on private blockchains that are fully controlled by the central bank itself. These blockchains are not open to public participation and offer a high level of security and governance.
  3. Public Blockchains: While less common, some central banks are considering public blockchains like Ethereum for their CBDCs. Public blockchains provide transparency and decentralization, but central banks would need to carefully address privacy and control issues.
  4. Hybrid Solutions: Central banks might also opt for hybrid solutions, combining aspects of both private and public blockchains to strike a balance between transparency, control, and security.

Prompt: what are the values of ethereum blockchain in case of cbdc’s that make it better for such usage?

  1. Smart Contracts: Ethereum’s ability to execute smart contracts enables automated and programmable transactions. Central banks can use smart contracts to define and automate various financial processes, such as interest payments, compliance checks, and regulatory requirements.
  2. Interoperability: Ethereum is compatible with various standards and can interact with other blockchain networks, allowing for cross-chain transactions and potential interconnectivity between different CBDC implementations. This can be crucial for cross-border transactions and international trade.
  3. Decentralization: While CBDCs are typically centralized, using Ethereum allows central banks to leverage the benefits of a public blockchain while still maintaining control. They can choose the degree of decentralization that aligns with their monetary policy and security requirements.
  4. Security: Ethereum has a robust security infrastructure, with a large and active developer community continuously improving the network. This can help central banks ensure the safety and resilience of their CBDCs.
  5. Transparency: Ethereum’s public blockchain offers a high level of transparency, making it easier for regulators and the public to audit transactions. This transparency can enhance trust and accountability in the issuance and management of CBDCs.
  6. Immutable Record: Ethereum’s blockchain provides an immutable record of all transactions, which can be valuable for auditing and dispute resolution. It ensures that transaction history cannot be altered or tampered with.
  7. Token Standards: Ethereum has well-established token standards like ERC-20 and ERC-721, making it easier to create and manage digital assets, including CBDCs and other financial instruments.

Prompt: what are the requirements for blockchain to be adapted in cbdc ?

  1. Security: Blockchain systems used for CBDCs must prioritize robust security measures to protect against cyberattacks, fraud, and unauthorized access. This includes encryption, key management, and secure coding practices.
  2. Scalability: The blockchain must be capable of handling a high volume of transactions to support the needs of the entire economy. Scalability solutions like sharding or layer 2 solutions may be necessary.
  3. Privacy and Confidentiality: Depending on the design of the CBDC, privacy considerations can vary. The blockchain should provide an appropriate level of privacy for transaction participants while ensuring compliance with anti-money laundering (AML) and know your customer (KYC) regulations.
  4. Interoperability: The ability to interact with other blockchains, financial systems, and payment networks is essential for cross-border transactions and financial market integration.
  5. Regulatory Compliance: The blockchain system must comply with relevant financial and data protection regulations in the jurisdiction where it operates. This includes adherence to AML, KYC, and other financial regulations.
  6. Governance Framework: Central banks should establish clear governance mechanisms for the blockchain network, including decision-making processes, upgrades, and protocol changes.
  7. Central Bank Control: The central bank must have control over the issuance, management, and distribution of the CBDC to maintain monetary policy authority.
  8. Resilience and Redundancy: CBDC blockchain networks should have redundancy and fail-safe mechanisms to ensure system uptime and resilience against potential disruptions.
  9. User Accessibility: The system should be user-friendly and accessible to a wide range of participants, including the general public, financial institutions, and businesses.
  10. Auditing and Transparency: The blockchain should offer transparency to allow regulatory authorities and auditors to monitor transactions and verify the integrity of the system.
  11. Cross-Border Functionality: If international transactions are a consideration, the CBDC blockchain should enable cross-border payments and adhere to international standards.
  12. Token Standards: If the CBDC is tokenized, adherence to established token standards (e.g., ERC-20 or similar) can simplify integration with existing blockchain and financial infrastructure.
  13. Digital Identity: Robust digital identity and authentication mechanisms are critical to ensure that participants in the CBDC system are appropriately identified and verified.
  14. User Support and Education: A well-planned strategy for user support and education is crucial for the successful adoption of CBDCs, especially when the technology is new to a significant portion of the population.
  15. Scalable Infrastructure: The underlying infrastructure, including servers, storage, and networking, should be scalable to accommodate increased demand as adoption grows.

Prompt: what impact on crypto may cbdc’s have?

  1. Increased Competition: CBDCs could pose competition to existing cryptocurrencies, particularly stablecoins like USDC and USDT. Users might find CBDCs more attractive due to their stability, backing by central banks, and government guarantees.
  2. Regulatory Pressure: As governments introduce CBDCs, they may intensify their regulatory focus on other cryptocurrencies. This could lead to increased scrutiny, reporting requirements, and potential restrictions on certain crypto activities.
  3. Privacy Concerns: CBDCs might raise privacy concerns since they are typically issued and monitored by central banks. Users who value privacy may gravitate toward more privacy-focused cryptocurrencies like Monero or Zcash.
  4. Cross-Border Transactions: CBDCs could simplify cross-border transactions and reduce the need for intermediary banks, which is a significant use case for cryptocurrencies. Crypto projects aimed at facilitating cross-border payments may face competition from CBDCs.
  5. Volatility Mitigation: CBDCs may attract individuals and businesses looking for a digital currency with the stability of fiat. This could lead to reduced interest in cryptocurrencies as a store of value or medium of exchange.
  6. Integration with Existing Financial Systems: CBDCs are likely to integrate more seamlessly with traditional financial systems and infrastructure, potentially reducing the need for intermediaries that cryptocurrencies aimed to disrupt.
  7. Innovation and Collaboration: Central banks may collaborate with the private sector and leverage blockchain technology to create CBDCs. This could lead to advancements in blockchain technology and innovations within the cryptocurrency space.
  8. User Adoption: If governments actively promote and encourage the use of CBDCs, they may become the preferred digital payment method for many, potentially diminishing the interest and adoption of other cryptocurrencies.
  9. CBDC-Backed Tokens: Some governments may explore using blockchain technology to issue tokens representing CBDCs on public blockchains. This could create a bridge between CBDCs and the broader crypto ecosystem.
  10. Diversification of Investment Portfolios: Investors and institutions may view CBDCs as another asset class in addition to traditional fiat and cryptocurrencies, diversifying their portfolios.

This post is pretty general and focused on the topics that intresed me so if u want to do some research on your own here are the useful links:

Bank of international settlements:

Central bank digital currencies – Executive Summary ->

https://www.bis.org/fsi/fsisummaries/cbdcs.htm

Lessons learnt on CBDCs ->

https://www.bis.org/publ/othp73.htm

you can also use search bar on bis website for more specific pdf’s.

CBDC’s tracker:

Self explaining – tracks where and on what stage cbdc are developed ->

https://www.atlanticcouncil.org /cbdctracker/

bridge-ing the crypto and cbdc:

use kozminski email as a “professional” email if regular one don’t work:

https://ripple.com/lp/cbdc-dea-whitepaper/?c1=GAW_SE_NW&source=CBDC_INTL&cr2=search__-__intl__-__non_-_brand__-__cbdc–general__-__exm&kw=central_bank_digital_currency_exm&cr5=676224205111&cr7=c&utm_source=google&utm_medium=cpc&utm_campaign=search__-__intl__-__non_-_brand__-__cbdc&utm_term=central_bank_digital_currency_exm&utm_content=general__-__exm&gclid=Cj0KCQjw4bipBhCyARIsAFsieCzZ8O–3MkMDw5rSaQapmCcG4TYdsLa9OnjA3HdXhNVzqX20l8DTZAaAvk5EALw_wcB

Japan’s central bank results on cbdc testing:

https://www.boj.or.jp/en/paym/digital/dig230529a.pdf

Poland’s central bank developed take on cbdc:

https://nbp.pl/en/payment-system/statistical-data/analyzes-and-studies/central-bank-digital-currency/

Tagged