Tech’s impact on democracy, legitimacy in systems, Ganz and perils of funding movements, misc tech review

Research / Theory

The Tech Coup: How to Save Democracy From Silicon Valley (Marietje Schaake)

• Primary argument is that technology companies are increasingly outside of the control of governance.
• Technology companies are rapidly moving into spheres of what was previously public good provision. This includes major cables and connects; starlink internet; etc.
• For AI/LLMs, technology companies are making enormous investments. This creates an incentive to recoup investments through future capture of value.
• Publicly funding AI models and other major platforms for public, ‘fair’ use is crucial.
  • The US is unlikely to do this, but the EU may.
  • Social media is inherently monopolistic. However, these new models do not have strong network effects.
  • The ‘marginal cost’ of production is quite low.
• Technology regulations can force immense shifts due to the incentives for these companies to not create different versions of the same software.
• Technology companies are adversarial towards regulation and governance, often flouting or evading rules. Uber is the most prominent example, but it is widespread.

Techno-optimism is only one side of the coin

• Technology is effect agnostic. Techno-optimism has become common since the dotcom boom and again during social media’s early impact on social revolutions. However, technology can be used to centralize power and reduce the quality of people’s lives. This point rhymes with Acemoglu and Johnson (2023) and Vitalik Buterin’s concern about AI being used to expand the capacity of highly centralized regimes. Even worse, AI does not have agency, meaning there is no ‘Human in the Loop (AI)’.
  • The sky is near, the emperor is everywhere. Vitalik Buterin uses a play on the Chinese proverb "天高皇帝远 ("tian gao huang di yuan"), "the sky is high, the emperor is far away." AI can massively increase the capacity of centralized states to exert control through semi-autonomous agents. What does counter-play (agency promoting AI uses) look like?

Expertise asymmetry

• Silicon valley understands and can explore uses of AI much more effectively than the masses or political representatives can foresee their uses.
  • This is a fundamental difference from decentralized protocols (email, rss, ssh) or social media (use cases generated by the users in a public ‘sandbox’).

“The Most Important Scarce Resource is Legitimacy” Legitimacy as social norms

• Vitalik Buterin summarizes the scarcity and value of legitimacy as a social technology.
• In decentralized coordination games, legitimacy is often king/queen. He (as may be expected) provides examples from crypto, where central actors attempt to defect (i.e. Justin Sun buying Steem as a company to capture STEEM the crypto-coin). In these cases, the legitimacy of the token itself was destroyed, so the community re-built a clone token minus the defecting actors, restoring legitimacy.
  • This can be seen by the enormous, multi-millon USD outlays daily by both ETH and BTC to ensure security and trustworthiness. This is far more than efficient, unless we assume that (aspirationally) iron-clad legitimacy is itself incredibly valuable.
• Legitimacy here is nearly (or exactly?) synonymous with what (Bicchieri 2006, 2016) terms ‘social norms’, being composed of both descriptive and injunctive norms. Vitalik notes that legitimacy has to do with second-order expectations about how all others in a coordination game will act.
• A useful breakdown of the paths to legitimacy follows (below quoted):
  • “Legitimacy by brute force: someone convinces everyone that they are powerful enough to impose their will and resisting them will be very hard. This drives most people to submit because each person expects that everyone else will be too scared to resist as well.
  • Legitimacy by continuity: if something was legitimate at time T, it is by default legitimate at time T+1.
  • Legitimacy by fairness: something can become legitimate because it satisfies an intuitive notion of fairness. See also: my post on credible neutrality, though note that this is not the only kind of fairness.
  • Legitimacy by process: if a process is legitimate, the outputs of that process gain legitimacy (eg. laws passed by democracies are sometimes described in this way).
  • Legitimacy by performance: if the outputs of a process lead to results that satisfy people, then that process can gain legitimacy (eg. successful dictatorships are sometimes described in this way).
  • Legitimacy by participation: if people participate in choosing an outcome, they are more likely to consider it legitimate. This is similar to fairness, but not quite: it rests on a psychological desire to be consistent with your previous actions.”

Funding has pernicious effects on movements

• Marshall Ganz argues that funding has robbed movements of their ability to attract a constituency.
  • Funding end-runs the requirement to build organizing capacity.
  • Mobilizing large groups of people, assuming money, is not difficult. However, this is not through the organizations capacity, and it is fundamentally not relational.
  • Ganz views civil society as at a nadir point in US history.
  • Civil society has become dependent on donor funding, rather than organizing the public
  • Note: The ability of one-to-many technology (social media) in mobilizing networks without organizational infrastructure or infrastructure may be driving this. Is civil society falling to McLuhan’s law (medium is the message).

Tooling

Quick python syntax review

Looping

• python can implement looping in many ways:

# for i in [...]
for i in [0, 1, 2, 3]: # this list is by item, and can be any datatype (i.e. strs are ok)
    # {do...}

# can also use range[int], which evenly distributes across the range.
# While not needed, np.arrange(from, to, increment size) allows for incrementing floats.
for i in range(3): # this will go 0-3, though two int arguments will start/stop at those values
    # {do...}

# while loops with an 'i'
i = 0
while i < 10:
    # {do...}

• Python loops have a number of advanced properties:

while True: # loops infinitely, requiring a break condition
    best_game = input("What is the best strategy game? ")
    if "starcraft" in best_game.lower():
        print("correct! ^_^")
        break # only stops prompting if starcraft is in answer.

• When looping over a string, python can loop over individual characters:

text= "IPSO FACT?"
for c in text:
    c_lower += c.lower() # += automatically concatenates strings. 
print(c_lower)

# more efficiently, you can use str.join
[c,lower() for c in text] # this is a list comprehension
# OR
(c,lower() for c in text) # this is a generator expression

Debugging

• Python natively includes ‘try’, where you can specify a piece of code to attempt and modify the output on failure:

try:
    num1 = int(input("integer to add #1: "))
    num2 = int(input("integer to add #2: "))
except ValueError
    print("... not an integer...")

SQLite review

• SQL databases are .db
  • to import a csv, use .csv mode (.mode csv >> .import {file}.csv {file})
  • .schema – this is similar to df.info() and df.describe in python’s pandas or str(df) in base R and glimpse(df) for tidyverse
• Basic commands:
  • SELECT {data} FROM {data object};
    • selects specified data from an object, can heavily modify intermediately
  • WHERE {variable} {logical condition}
    • This selects based on a
  • GROUP BY {variable};
    • Partitions based on unique values in a column BEFORE applying aggregating functions
    • HAVING applies filtering based on the GROUP BY function. WHERE, in contrast, only applies before aggregation
  • CAST({data} AS {NEW_TYPE});
    • recasts type
  • ROUND({data}, {#decimal places});
    • rounds
  • CASE WHEN "{variable}" {logical statement} THEN {DO} ELSE {ALT DO} END;
    • conditional statements, add a + at the end to add more conditional DO statements
  • LIKE is akin to regex searching, where we look for pattern matching especially in WHERE. % matches any number of characters, and _ is a wildcard. These can be mixed, for instance, you can specify any character, followed by obb and any ending as SELECT * from people WHERE name LIKE '_obb%', which
• Linking tables:
  • Linking tables is often more efficient than merging them.
  • SQL supports primary and foreign id.
  • The foreign id links to another table, while the primary is for that table.
  • The use of ids is to limit size in memory (ids are integers, not strings). Referencing relates in modular calls.
  • These can be nested so you have a pointer going ‘traversing’ multiple layers (a.item where id = b.a_id, where b.id = c.b_id… etc.).
• Indexing for faster searches
  • SQL allows for creation of an Index (CREATE INDEX foo ON fie)
  • This is a b-tree, not a dictionary. SQL generally does not support dictionary lookup.
  • Limits the time to go from the ‘top’ of the tree to ‘bottom’ leaf row.

Internet plumbing

• Internet Protocol (IP): a 4 or 6 part address for a server. There is a sender and receiver among the metadata. These are passed between ‘hops’
• TCP (Transmission Control Protocol) is a protocol which add information about the source port and destination port (e.g. 443 is the port for https, others for pop3 email etc.)
• DNS (Domain Name Service) converts domain name (google.com) to an IP. Your computer will query DNS server(s) to retrieve the IP of a domain name you request. Once it knows, it uses TCP/IP to communicate.
• DHCP (Dynamic Host Configuration Protocol) is how your computer …
  • ‘gets’ its IP address assigned if it has not been used for some amount of time. Necessary bc manual IP is not scalable.
  • ‘gets’ its best DNS server to contact as a ‘default gateway’ or router