The recording link is the Afternoon TEE Party side event during DevCon: https://m.youtube.com/watch?v=2XBQsvLVzyc

00:55 | Decentralized Confidential Computing (DeCC) on iExec
10:24 | BlockVote proof of concept overview on Oasis
31:25 | Explain TEEs to me like I am a child
39:24 | Hardware trust assumption critiques of TEEs
46:50 | AI and TEEs vs. cryptographic guarantees (e.g. ZKP)
51:35 | TEEs and ML experiments & limitations

Here are my key takeaways

Why is everyone talking about TEE? It’s simple. The hardware is easier to use now.

• The hardware capabilities for writing an application for TEE are much different today than in 2015 vs 2017 vs 2022.

• TEEs are usable now but need a combination of cryptographic guarantees for data-in-use protection (hardware + math is better)

• TEEs can provide AI with low securities guarantees today, whereas cryptographic guarantees (e.g., ZKP) are very safe but have yet to work in an actual application.

• Concerns about trusting the hardware manufacturer are still prevalent (crypto hate(d) TEEs)

• Future TEE predictions? Abstract TEE to software level and open source TEE hardware manufacturing.

BlockVote is a voting platform built on Oasis Sapphire, a privacy-focused EVM that allows encrypted transactions and storage.

BlockVote provides the option for:

• Gasless voting (eliminate the need for voters to pay gas fees), 

• Customizable voting (settings for token weighting, privacy levels, and eligible voters) and 

• Privacy options to choose how much information is revealed, from fully private to fully transparent votes.

BlockVote is a proof-of-concept private voting dApp only live on the test net.

iExec provides decentralized, blockchain-based cloud computing services. The project's core offering includes confidential computing. 

• Developers deploying applications on iExec have on-demand access to a network of TEE machines that protects data during processing.

• iExec is part of the Decentralized Confidential Computing alliance, a blockchain network coalition encompassing various encryption technologies and solutions (e.g., TEE, ZKP, MPC).

Text agents have been around since the beginning of AI. Early text agents could interact with their environment and perform actions based on language input. ELIZA is a rule-based chatbot developed in the 1960s and marked an early milestone in the development of text agents. ELIZA is an example of the AI paradigm called symbolic AI agent, a paradigm spanning from 1960 through the 1990s. Like sophisticated reasoning engines, symbolic AI agents program all the rules to interact with different environments. ELIZA relied entirely on pre-programmed rules and used simple pattern-matching to create seemingly human-like conversational ability. ELIZA was limited by domain specificity and manual design and could only work up to simple domains. The lack of computational power during this era also prohibited the ability to do anything substantial to advance AI. While groundbreaking, these systems ultimately highlighted the need for more adaptable and intelligent approaches.

The emergence of machine learning and reinforcement learning (RL) ushered in the next phase for text agents starting from the mid-2000s to the present day. RL agents learn through interaction with their environments, receiving rewards for desirable actions and adjusting their behavior accordingly. This data-driven approach enabled agents to show greater flexibility and adaptability than their rule-based predecessors. However, pre-LLM RL agents were still limited by their domain specificity and reliance on clearly defined reward signals. Each new task often required extensive training from scratch, hindering their generalizability and scalability.

Powerful large language models (LLMs) exist today, and we now have reasoning agents that can understand complex instructions, plan actions, and adapt to circumstances. The advent of LLMs revolutionized the AI field, giving rise to LLM-powered reasoning agents. One example of an LLM agent is GPT-3, a machine learning model that uses deep learning to generate text that can seem like a human wrote it. These agents leverage the vast language understanding, reasoning, and knowledge inherent in LLMs to overcome the limitations of previous approaches. 

Another LLM agent application I found exciting in the lecture is SWE-Bench, an LLM agent for code programming. In SWE-Bench (2023), presents a text agent with a GitHub repository and a specific issue. The agent needs to analyze the codebase, understand the problem, and generate a file diff that resolves the issue, just like a human software engineer would. This text agent goes beyond simply generating code snippets; it needs to interact with the code, run tests, and iterate on its solution.

Despite these advancements, LLM-powered reasoning agents still face challenges in developing intelligent text agents. One active research area is long-term memory, enabling agents to retain knowledge and learn from past experiences. Developing new benchmarks and evaluation metrics that go beyond simple task completion and emphasize robustness and human-like interaction is also crucial. As the field continues to evolve, the future holds exciting possibilities for text agents to become indispensable digital automation tools and problem-solving tools.

***Author's note: This article submission is to satisfy the written assignment for the Fall 2024 MOOC course, Learning Language Model Agents, facilitated by the UC Berkeley Center on Responsible Decentralized Intelligence (RDI) .***

What is SheFi?

SheFi is a globally diverse learning community and professional network. The organization's mission is to empower professional women and non-binary people to reinvent their careers in Web3. Throughout the year, SheFi offers an eight-week Web3 learning program organized into "seasons." There is also a Discord server for questions and community support from current and former season scholars.

How is the learning program structured?

The program is a series of one-hour, eight-week live classes that broadly cover crypto and Web3 basics. Topics include understanding blockchain, crypto wallets, layer 2 blockchains, decentralized exchanges, liquid staking, NFTs, Web3 social, governance, public goods, ENS, AI, consumer crypto, zero-knowledge proof, and more. Every class includes a live Q&A with SheFI founder Maggie Love. The program also involves on-chain 'quests' to get hands-on experience with topics covered in class.

How was my SheFi experience?

Overall, I really dig learning crypto this way! It is a comprehensive approach to learning Web3 basics across the various sectors of the industry.

Pros

The live sessions are fun and approachable. The learning materials include guides, recorded video tutorials, and other supplemental content beyond the live classes. The content is easily digestible, and the scholars help each other understand it. The learning examples also center on women's experiences. Going into the course, I was skeptical about how a female-oriented online learning cohort would be structured differently than a general learning experience, but it does help to create a safe and fun environment to learn and understand Web3.

The on-chain "quests" activities make the learning materials more engaging. They also give learners on-chain practice, the best way to learn Web3. SheFi developed its wallet, which scholars can download to practice transferring and swapping crypto.

Cons

The course does emphasize a "wealth-building" perspective. What that looks like in the learning materials are deep dives into liquid staking, farming, and other DeFi opportunities. There was one deep dive into zero-knowledge proofs as part of the Layer 2 unit, but I would have liked to have seen more cryptography deep dives. Small disclaimer: I am enamored by blockchain technology's cryptography and computer science advancements, but I never really cared for the investment aspects of Web3. It was nice, however, to learn the difference between a derivative and a call option.

Should you join SheFi?

Yes! I recommend anyone, from a newbie to an advanced learner, apply for SheFi. I know blockchain technology and have learned something new from each lecture. Plus, the live sessions sometimes featured guest speakers who shared upcoming changes to their protocol. I also recommend applying for a scholarship to cover tuition if your workplace can't afford it. I was awarded a scholarship provided by Celo this season.

In 2021, the Pew Center research survey measured Black and Hispanic folks as the highest leading demographics in cryptocurrency adoption in the United States. Driving this adoption rate could also be explained by central finance narratives for lack of participation from communities of color. Such as distrust in traditional financial institutions, the historical abuse, racism in interacting with them, or the simple fact that lenders lend to their network is limited to reflect the life experiences represented in your employee make-up. The problematic relationship between Black households and financial institutions alone has quantifiable consequences; the FDIC 2017 national survey determined 47 percent of Black households are unbanked or underbanked.

Distrust and poor user experiences in financial institutions could explain some Black and Brown user adoption. Still, another contributing factor to this ease and rapid user adoption is that, inherently, such communities already have a collective mindset. When examining inherent bias or traditions in social groups, colonizing cultures is used to an extractive relationship. Meanwhile, indigenous societies existed for thousands of years to reach community actualization, as explained from this First Nation perspective. In present-day crypto land, this stored memory among marginalized folks translates to sharing a network across distributed nodes or pooling resources together for a DAO or DeFi exchange because there is already an emphasis in the physical world on community care. Transformation with DLTs could perhaps draw in Black and Brown innovators who can inherently embrace these collective community mindsets and are open to understanding technology.

In 2018 I participated in a hackathon partially sponsored by Holochain to expose women to blockchain technologies. I still remember the conversation and lightbulb moment with co-founder Arthur Brock. The idea the team was playing around with was a proof of employment history where time spent at a company was stored on-chain, dates, company name, and position title(s). What made it a blockchain use case was that the applicant user could share the public hash of their experiences. Thus, pseudo-anonymously prove their work history without over-sharing personal information of a full-blown employer verification early in the applicant process.

Why would this matter? From the team's perspective, giving so much data, even from a first or last name, could potentially introduce unwanted bias into the candidacy pursuit. For instance, you might have to overshare a different last name used at a previous job. For a woman in the United States, that could flag questions about whether she was recently married or divorced. A first name change could also give away contradicting gender impressions or force the applicant to share their dead name. Someone with a foreign-sounding name could introduce questions about immigration status or racial assumptions. All this is presented by the data point of a first and last name when as a candidate, verification of work history should focus on confirming work occurring at a company. At least, that was what our pitch argued. The technology is neutral and almost always the easy part. In this case, blockchain and the dApps platforms were already there. The ideas of what to build on came from diverse women with various biased experiences and collective mindsets when searching for meaningful tech work.

There is an explosion of beautiful projects with traces of inclusion in this space. Blockchain is paradigm-shifting in technology solutions but not revolutionary on its own. When Bitcoin launched, there was never a goal for financial inclusion. There is only so much to expect from the launch of a successful combination of game theories and computational equations with cryptographic features. Bitcoin's white paper proved that these features could work for a decentralized, immutable ledger. But, it never claimed to be a project for financial inclusion. Yet, I've observed that folks who have spent time trying to solve these wicked problems could connect the dots given the underlying blockchain technology principles. I am hopeful that entrepreneurs will leapfrog innovating in centralized finance and web2 fintech and instead build meaningful, community-altering web3 solutions for womxn, womxn of color, Black crypto, and beyond. 

Human-Computer Interaction (HCI)

A research focus developed in the 1980s when the personal computer became more readily available, Human-Computer Interaction (HCI) observes the interconnectedness of three elements: the individual or people, technological use or device, and the behavior between the two entities. It is also a multidisciplinary subject that focuses on computer design and user experience to understand and facilitate better interactions between users and machines. 

There are sub-topics within the HCI umbrella, particularly when you layer in human identities with socio-economic, disability, ethnic, or gender studies. I argue, however, that this is adjacent to but separate from conversations on diversity and technology. Diversity topics include studies such as machine learning algorithm bias towards people of color. In short, when you lack representation from different life experiences, your technology will always have a deficit in addressing all global users. These are critical discussions, but HCI takes another step beyond pointing out the source of trends by understanding why the trends exist.

Postcolonial Computing

Postcolonial computing is a sub-topic under the HCI umbrella. Therefore, it was the appropriate lens through which I conducted my blockchain research within an international setting. Keeping to a simple Wikipedia definition, postcolonialism is "inspired by the academic study of the cultural, political, and economic legacy of colonialism and imperialism, focusing on the impact of human control and exploitation of colonized people and their lands." Thus, these critical theories of the peoples, societies, and diasporas inspire postcolonial computing and how it affects technology and technological development. 

This framework is crucial in moving beyond models recognized as "tech for good," "ICT4D," and "digital development." Think back to the US Digital Service origin story and the failed first attempt at the 2014 Healthcare.gov website. Yes, that 2005 - 2017 phase of taking x industry for social good (triple bottom line, B-corps, etc.). Good intentions but flawed in approach. Taking a page from nonprofit/NGO critical studies of the 1970s, it's hard to create generational change with an inherent power imbalance in the relationship.

Tech for good was my starting point. As I studied the Information and Communication Technologies (ICT) for Development (ICT4D) movement, I was learning their design principles and feeling uncomfortable. ICT4D is a mix of user-centered design using digital solutions to accelerate communities in their social and economic development. TL: DR, the global north is sorting out how to interact with the global south. Around the same time, I dedicated myself to unlearning euro-western-centric methodologies and felt that the ICT4D framework needed to be revised. Nevertheless, I kept digging until I found theories that fit my reality as a foreigner living and working in a South American country.

Postcolonial computing theory highlights

In comes postcolonial computing, the first framework that really 'clicked' with me. It considers the imbalances in power, history, identity, and the theory of knowledge realities. Kavita Philip describes postcolonial computing as "a way of asking questions, a mode of investigating, and a form of conversation." Some key points I found most helpful in understanding postcolonial computing as:

• Recognizes the emergence of hybrid knowledge practice across communities, both ancestral innovations and technologically driven;

• Technological artifacts embody the politics of their design and development, and separating the two is unfeasible; and

• Questioning who is "us" and "them" in these conversations, thus reimagining the global distributions of the knowledge economy and computational labor.

This approach doesn't center or lament any one party, the oppressor or the oppressed. It also encourages questioning the context of how and for whom new technologies are developed. It considers the histories that subconsciously drive our interactions within the same design framework. Although annoying, rather than provide all the answers or a "right" path, postcolonial computing prefers a critical awareness of our actions. 

Frustrations in Web3

This awareness of postcolonial computing was vital to unlocking my mental block when I wanted to conduct a blockchain-focused research project. In 2017 I observed that countries in the global south were somehow lab rat experiments for half-baked distributed ledger technology projects. Most DLT design ideas could work better with a centralized database approach instead. The DLT justification is a constant struggle I still observe in the space, the balance of innovating with this fantastic technology design at the risk of experimenting with already vulnerable communities. Taking advantage of vulnerable communities is becoming apparent in mainstream media. In his article for Vice, Luke Ottenhof vents about Crypto-Colonists most recently, but these kinds of criticisms go back to the Rohingya Project.

Discern for yourself

HCI and postcolonial computing are my favorite topics to discuss, so expect more musings around this topic and specific applications. Here are links to some valuable primers for additional reading on postcolonial computing.

(1) Postcolonial Computing: A Tactical Survey. This is a good overview and valuable examples:  https://www.academia.edu/8605271/Postcolonial_Computing_A_Tactical_Survey

(2) Postcolonial Computing: A Lens on Design and Development. This one is a little denser, but I like its critical argument on the ICT4D movement:  https://www.wtf.tw/ref/irani_postcolonial_computing.pdf

Happy discerning!

Why 'discernment'?

I love the paradigm shift brought on by blockchain technologies so much that I can be critical of their application. My first project deep dive into the application of blockchain technologies was during my 2017 -18 Fulbright grant, where I studied government adoption of blockchain technology. During the ICO frivolity, I was in the middle of sorting through why we even use blockchain as the underlying tech stack. I remember the excitement around the Rohingya Project and experimentation in issuing land titles in Georgia. Intelligent contracts were a favorite buzzword but with limited iterations and field examples. In all this flurry of activity (did the eruption ever stop?) I grounded myself in the principles of distributed ledger technology and always asked whether a distributed system across nodes was necessary. Discernment became my favorite reminder throughout my web3 journey, and that habit never stopped. I’ve grown an aptitude for stepping back and being critical of my decisions and life approaches.

What readers can expect

I will explore topics contextualizing Human-Computer Interaction (HCI) theory disciplines, particularly postcolonial computing, and computer-supported cooperative work (CSCW) concepts. I am driven by the intersection of government and blockchain technologies (monetary policy, stablecoins, CBDCs, etc.) and civic endeavors that address gender and racial inclusion in the technology and finance industry. I’m a policy nerd, but I value demystifying the swirling blockchain universes. I also plan to share my accounts of my experiences in web3 and connect the dots to broader themes of social justice and community.

Let me know how you like using Paragraph! I was on autopilot setting up a blog on either Substack or WordPress route out of convenience and familiarity. I realized I wanted to align my web3 values and challenge myself to learn something new.