Stocahstic Contemplation

It's incredibly insulting that economist and media pundits are calling the current economic stress of 70% of the population the "vibecession".

Since names matter, we should call it what it is. The middle class squeeze. As capital owner interests squeeze ever more profits out of a middle class that can ill afford it at this point.

The tipping point already happened long ago in the early 1970's. That every subsequent generation is going to be worse off economically than the generation before it and we're 4 generations in now.

To say it's all in your head because the top 20% are doing OK and the top 1% actually is seeing 60% gains is really disingenuous and does not match the lived reality of the 70% majority.

The macro economic numbers are good, the economy didn't shrink/contract. So therefor we are not in a recession according to the macro numbers. That is meaningless to 70% of the US population living paycheck to paycheck who are under major economic stress having to make ugly decisions about debt vs children vs education vs food vs housing vs medical.

It doesn't matter that inflation has come down. The already applied massive increases in prices for medical, child care, education, groceries, gas, and rent already tipped this 70% of the US into often being insolvent. This group can't cover all the cost of living multiple months of the year not just once and a while.

Driving an economy on macro numbers is incredibly myopic if they don't represent the majority of your population. Macro economic numbers are incredibly skewed due to 80% of wealth of the US belonging to the top 20% of earners in the US. So the top 20% distort macro numbers making it look like average earnings are in the $70k/year range. Looking at that number misses the massive opportunity inequality leading to 70% of the US population living in scarcity at sub $55k annual earnings pay while the top 20% live at $120k+ annual earnings PLUS the ability to invest.

How do I plan a session?

The 5x5 method

http://www.critical-hits.com/blog/2011/03/04/the-5x5-method-compendium/ Come up with 4-5 events/plot points that the players can encounter.  This can be as simple few phrases on an index card like

Encounter a Griffin trapped in a Net.

The Orc tribe that set the net

The Secret mine behind the orcs

The Drake in the secret mine

The entrance to the Under dark

Then as you run the campaign work theses into the narrative as you go along. The gist of the encounter remains though the circumstances may change. As you develop new quests or story threads write a grid with the threads across the top and your quest/major events along the edge. All story threads need to intersect with one of your major encounters or plot points at least in one location. How do I tell a story using the medium of RPGs?  

Story Structure

Stasis

Trigger

The quest

Surprise

Critical choice

Climax

Reversal

Resolution

The 8 point story arc method is generally a good starting point.  You don't need all the steps you can jump through just the Trigger, Quest,  Surprise/Obstical, Critical Choice, Climax, and Resolution.

Story Hooks

DnD is about giving dangling lines of a story and having players finish that plot.  So to give use the 8 point arc as a method you just drop Triggers about and it's up to characters to follow those triggers.

DnD isn't a novel

RPGS are not like books or novels.  RPGs are collaborative story telling where the group of you are telling a story together.   It's more like improv that any other medium.  This means players have to be vested and engaging with the world.    Otherwise if things only happen to the players but the players don't proactively do things on their own that’s a big problem for the any role playing game.

Narrative control

You don't have direct narrative control over the players actions.  What you can do is given them options to see or do or try.  Which is prompting.  Here's something interesting you see.  Have you considered this?  Make a roll for history maybe your character knows something of the history of this place?

A far superior situation is that the players themselves explore the world on with their own direction, merits and skills.  This is a more dynamic, driven by character needs and interests. 

How to come up with interesting characters.

IE don't be the dull one who has the personality of a stale slice of bread. DnD isn't about being told a story passively by the GM. At its heart you the players are telling the story of your characters.  Their actions, their great works and their mis-deeds.  Trials and tribulations of their daily lives and the adventure that drives them.

You don't have to come up with a persona complete with funny voice, but you do have to come up with how your character interacts with the world. What drives them? Why do they adventure? What are their goals? What do they believe? Note, you also have to work in the context of a team since DnD is an ensemble of characters in a party working together. So be vested and interested in other players stories and success and you'll have a much better time playing DnD. So don't be going and nullifying each others actions all the time. Let the interesting thing happen so long as collaboration is happening. Player vs Player Conflict. If someone is just being a jerk and trolling everyone by causing trouble then call them out on it and the GM needs to step in and resolved conflict. There is no this jerk behavior is justified based on my character would do that. The player is always in charge of what happens with the characters behavior and that's on the player in all cases when they decide to be anti-party or anti-player. The game mechanics BTW are not tools to resolve player conflict, you have to use your words human to human and to do that. What do I think about heroic action?  Am I a hero? Who do I look up to?

Am I merciful? Do I kill that surrendered goblin?

Am I religious? What do I believe in as far as the gods?

Why do I adventure?

What are my short term goals?

Who are your parents and mentors? Round it out with Quirks and Mannerisms.

What materials should I have as a DM?

Dice

Some tokens or miniatures. I recommend paper ones initially they pack flat and look nice.

An erasable battle mat with markers. Even if you only intend to do theater of the mind drawing a scene out for players is a powerful tool that avoids confusion.

Food and drink, you're going to be talking and thinking for 3 hours so bring something to eat and drink.

A notebook and pencil for taking notes. You just need some high-lites to jog your memory of the last session and NPC names/notes so you can keep them consistent.

How to build fights. I don't understand this crazy challenge rating system?

The quick answer is you don't need to understand the Challenge Rating System. How do I balance an encounter to be challenging but not kill the players follows this concept of Action Economy. Who has more actions in a round. If the players have more actions they have a significant advantage and if the monsters do they have the advantage.

The side with more actions can do more to the side that doesn't and can recover from setbacks easier. A good DnD supplement that uses this on the fly style encounter building is Forge of Foes from Sly Flourish. And it has a good easy rule of thumb difficulty metric.

How many HP should monsters have?

Easier: monster HP < Heroes per monster. Harder: monster HP > Heroes per monster.

A Solo Boss 4x to 6x what an individual fighter type character has.

Where should their roll bonuses be at?

Generally as good as the characters

Saves

Pick 2 stats, and it gets the same proficiency bonus are characters +1 OR +2, otherwise it's a strait roll with +1 or +2

Damage

Look at what the players are doing as damage. Easier: damage < heroes, Harder: damage > heroes.

Special Abilities

Does this ability degrade their ability to fight beyond just damage? If yes then this is a harder encounter because regardless of all other prevailing balance points because the longer the fight goes on the worse it gets for player characters.

Rewards

Far too long, but AI is a complex topic.

What are the real down to brass tacks things that need to happen in the discussion of AI?

AI is a complex topic so just noodling over generative AI or the unethical acquisition of training data is incomplete. You could gather the data in a completely ethical way and still have an AI that does horrible things like prevent women from getting engineering jobs. That's a real thing, Google AI resume screener filtered out resumes from women engineers for nearly a year before legal challenges showed a pattern of discriminating against women engineers.

So the next time you discuss AI here's the core regulatory concerns that need to get addressed.

First, AI is already here. It's not some far flung killer robot Skynet threat. AI is already embedded into many systems you don't even know determining if you get to live a good life.

Finance - Do you get a loan? Housing - Do you get to live in here based on the finance question? Medicine - Do you get treatment? Are your correctly diagnosed? Does your insurance pay or delay payment? How much does your insurance cover? Do you get a job? AI resume screeners are all over the place. Transportation - How much does it cost for gas, for an uber, for a car? Cost of Goods and Services - How much is this service or good?

AI is answering all these questions right now based on a set of priorities and goals set by companies and executives. What is good for corporations is clearly not the same as what is good for people and society in general. One only has to look at the pharma industry and insurance industry to see that.

The main issue is Transparency of AI currently. It's a black box with no regulation on how it's trained, created, distributed, and used. You are participating in using AI right now by using social media Apps along with 80% of the US population. AI owners would like you to believe they shouldn't be regulated. However when you have near 100% of your population being affected by AI decision making, that's well worth some regulation. This is the time for artists, writers, and normal people to insist on some regulatory rules of the road written for consumers not corporations.

The main points of AI regulation How and where did the company get the training data? There needs to be clear bright line legal definitions of what is ethical data harvesting. How you can do it. When you can do it. Where you can get data. Is consent involved. Follow the rules and you are safe. Don't follow rules, face massive penalties.

All the training sets must be made publicly inspectable by experts in a controlled environments. This is the only viable way the public is protected from bad faith drag net collection of IP protected work or private data. Why in controlled environments? Because by their nature some training data like cancer data is private information that should not be accessed without strong controls of who can see this data. Data must be able to be purged from model and training set. AI specialist don't deal with this currently because they aren't required too. They absolutely should be building this into the technology however. There needs to be a legal framework just like your credit score to get AI to forget data. And strong penalties for not doing so. There isn't a clean way to do this with current AI. The technology will be made if the regulation exists.

Was this data gathered in an ethical way? Using third world labor to scrub your data of the worst of the internet is not ethical and should be illegal and penalized.

Ill gotten training material that has other's IP should not be allowed to be me ac culpa-ed after the fact with no penalty. Companies and individuals should be forced to disgorge profits as well as play penalties for the practice. Just like any other form of IP theft. And like any law ignorance of the law or the IP is not a defense.

Owners of AI should be legally responsible for decisions made by that AI. The users of AI are also legally responsible for it's use. AI owners cannot feign ignorance of what the AI does. AIs are trained to be fit for a particular purpose by their owners baking in any bias and goals the owners have. The AI is a proxy for the decision making of the humans that designed it. So regulation should penalized owners for the bad decisions their AI's make just like any other product safety concern.

The AI company and the user/company using the AI cannot whitewash their decisions behind AI. Both the AI owners and AI users are at fault when issues arise. The user of the AI didn't have to use that product. Users using that AI leading to harmful or illegal results are not indemnified from responsibility. Even in the area of categorization or recommendation AI which are forms of unsupervised learning. The AI expert is choosing directives and goals for the AI to achieve like efficient protein folding it's not that "any" outcome the AI finds is acceptable. Who can sue and seek remedies?

That should be anyone. Not just governments or regulatory agencies. AI owners and users should face unlimited liability like any product/tool that is use in determining the quality of life of humans.

Eliminating negative human bias from AI Along the same lines of legal responsibility. Human training data has human bias baked into it and it is hard to remove. As does the methodology and methods of training the AI. That doesn't mean AI owners get a free pass on resolving that bias. So when AIs trained on immoral and illegal behavior are identified to do that bad thing the AI owner should have a limited time shut down the AI and resolve the issue and cannot be allowed to continue the AI operations in the biased state. Only then will the right incentives be in place to create ethical AI.

The full bevy of per instance and daily penalties should be applied in for Equal Opportunity and Rights violations; Financial and Medical privacy violations; and Insider trading to the company involved. And the AI itself as a technology must make this review and enforcement easy or the AI cannot be launched. This often heard refrain of "it's too hard" or "too costly" is bullshit. Regulation just needs to make it so costly to fail on transparency that AI companies can't choose to skip transparency.

After reading all these articles about using Amazon Dash buttons as generic wifi enabled buttons you can hack for your own purposes,  I finally bought some when Amazon Prime Day had a sale on dash buttons for $0.99. 

This code looks simple: install python 2.6, install scapy packet sniffer library, and we are off to the races.  BTW scapy needs about 6 other libraries so read the instructions on the scapy site for Windows installation carefully.

The original code:

from scapy.all import *

def arp_display(pkt):  if pkt[ARP].op == 1: #who-has (request)    if pkt[ARP].psrc == '0.0.0.0': # ARP Probe      print "ARP Probe from: " + pkt[ARP].hwsrc

print sniff(prn=arp_display, filter="arp", store=0, count=10)

The code given works perfectly on a Mac or Linux box but not Windows.  Because of course it doesn’t; why would it.  Very disappointed.

C:\dash>python scan.py WARNING: No route found for IPv6 destination :: (no default route?) WARNING: can't import layer ipsec: 'module' object has no attribute 'IPPROTO_AH' Traceback (most recent call last):  File "scan.py", line 8, in <module>    print sniff(prn=arp_display, filter="arp", store=0, count=10)  File "C:\Python26\lib\site-packages\scapy\arch\windows\__init__.py", line 523, in sniff    r = prn(p)  File "scan.py", line 4, in arp_display    if pkt[ARP].op == 1: #who-has (request)  File "C:\Python26\lib\site-packages\scapy\packet.py", line 817, in __getitem__    raise IndexError("Layer [%s] not found" % lname) IndexError: Layer [ARP] not found WARNING: __del__: don't know how to close the file descriptor. Bugs ahead ! Please report this bug.

So after much scouring of the interwebs, here’s how you fix these errors.

First, to fix IPPROTO_AH not being defined on Windows installations of Scapy.

Go into ipsec.py in your python2.6 installation and add the following lines after the import statements.

if not hasattr(socket, "IPPROTO_ESP"):    socket.IPPROTO_ESP = 50 if not hasattr(socket, "IPPROTO_AH"):    socket.IPPROTO_AH = 51

Second, we need to modify the code provided because the Windows version of Scapy will return empty packets on the sniff call.  This causes the error: IndexError: Layer [ARP] not found because empty packets don’t have an ARP layer.  The sniff filter=“arp” parameter was suppose to filter down to only packets with the ARP layer but that obviously doesn’t work. 

The “prn” function we pass into sniff needs to handle this empty packet case.

So here’s the modified code for the scan of mac addresses:

from scapy.all import *

def arp_display(pkt):  if ARP in pkt and pkt[ARP].op == 1:    print "ARP Probe from: " + pkt[ARP].hwsrc

sniff(filter="arp", store=0, count=0, prn= arp_display)

So what does this fix?

The Windows version of scapy doesn’t show the psrc ip address as “0.0.0.0″.  It actually gives the DHCP ip the button bound to.  So that check had to be removed.

Since we get empty packets through the filter we need to check if the ARP layer exists in the packet.

The count=0 sets the sniff to continue to scan package infinitely until Ctrl-C.  Setting this to 10 packets is much to short a time in practice for the dash button to ack the network and have it show.

We don’t need to print the results of the sniff since we are printing data in the prn callback function.

If you want to see the full packet replace the print statement with:

print pkt.show()

If you have a network like mine there will be more devices than the dash button sending ARP packets. So you’ll need to watch for the mac address that’s best correlates to when you pressed your amazon dash button.

So once you have your mac address you can now look for it connecting.

Final code:

from scapy.all import * import time

starttime = -1

def arp_display(pkt):  global starttime  if ARP in pkt and pkt[ARP].hwsrc == "80:3f:5d:9a:60:60":    if (starttime == -1):      print "My Gatoraid button connected: " + pkt[ARP].hwsrc      starttime = time.time();    elif (time.time() - starttime > 4):      starttime = -1

sniff(filter="arp", store=0, count=0, prn= arp_display)

In action:

C:\dash>python scan2.py WARNING: No route found for IPv6 destination :: (no default route?) My Gatoraid button connected: 80:3f:5d:9a:60:61

I do not get single connect packets using this method so I needed to add a back off timer to ignore further reported packets in certain period. Have fun hacking dash buttons on Windows.