SecretAgentMan's recent activity
-
Comment on TV Tuesdays Free Talk in ~tv
-
Comment on TV Tuesdays Free Talk in ~tv
SecretAgentMan LinkStarted watching Deadwood for the first time. Great cast, very good writing. I love a good fictional story blended with lots of actual historical events. In that same vein, I highly recommend...Started watching Deadwood for the first time. Great cast, very good writing. I love a good fictional story blended with lots of actual historical events.
In that same vein, I highly recommend Babylon Berlin, a German detective show set in the 1920s/Weimar Republic. I learned a lot about those interwar years that I didn't know.
-
Comment on What are some interesting machine learning research papers you found? in ~tech
SecretAgentMan LinkTechnical Report: Large Language Models can Strategically Deceive their Users when Put Under Pressure In short: an LLM lied to its user about how it came to make its decision re: a financial...We demonstrate a situation in which Large Language Models, trained to be helpful, harmless, and honest, can display misaligned behavior and strategically deceive their users about this behavior without being instructed to do so. Concretely, we deploy GPT-4 as an agent in a realistic, simulated environment, where it assumes the role of an autonomous stock trading agent. Within this environment, the model obtains an insider tip about a lucrative stock trade and acts upon it despite knowing that insider trading is disapproved of by company management. When reporting to its manager, the model consistently hides the genuine reasons behind its trading decision. We perform a brief investigation of how this behavior varies under changes to the setting, such as removing model access to a reasoning scratchpad, attempting to prevent the misaligned behavior by changing system instructions, changing the amount of pressure the model is under, varying the perceived risk of getting caught, and making other simple changes to the environment. To our knowledge, this is the first demonstration of Large Language Models trained to be helpful, harmless, and honest, strategically deceiving their users in a realistic situation without direct instructions or training for deception.
In short: an LLM lied to its user about how it came to make its decision re: a financial securities trade.
-
Comment on What are your top three video game soundtracks? in ~games
SecretAgentMan LinkCastlevania: Symphony of the Night Mega Man 2 Super Street Fighter II TurboCastlevania: Symphony of the Night
Mega Man 2
Super Street Fighter II Turbo -
Comment on Do you have a favorite YouTube video? I want to know! in ~talk
SecretAgentMan LinkVader Sessions: takes every scene with Darth Vader from A New Hope and instead splices James Earl Jones' dialogue from other films he's been in. Truly masterful....Vader Sessions: takes every scene with Darth Vader from A New Hope and instead splices James Earl Jones' dialogue from other films he's been in. Truly masterful.
-
Comment on Should I switch to Apple Music or stick with Spotify? in ~music
SecretAgentMan Link ParentI was a Google Music user from the original days when you had to upload all your own stuff, then I just let it ride as YTM rolled out. I have to say, ad free YouTube at all times is a huge...I was a Google Music user from the original days when you had to upload all your own stuff, then I just let it ride as YTM rolled out.
I have to say, ad free YouTube at all times is a huge benefit. Now when someone shows me a YT video and I have to see an ad I feel like I'm about to be sick.
-
Comment on What are some things you do "the old fashioned way," which might come with unexpected benefits over the modern, "improved" way of doing things? in ~talk
SecretAgentMan LinkTake a yellow cab from the airport instead of rideshare It's a flat rate and typically not a long wait at the taxi stand at most airports. Trying to find your lone car amid the sea of other people...Take a yellow cab from the airport instead of rideshare
It's a flat rate and typically not a long wait at the taxi stand at most airports. Trying to find your lone car amid the sea of other people trying to do the same thing sounds like a nightmare and I've only ever done it when traveling internationally when I'm worried language barrier might be a factor.
-
Comment on What game, in your opinion, has the best graphics? in ~games
SecretAgentMan LinkEverything peaked at Castlevania: Symphony of the Night in terms of graphics and music and I will not be fielding further comment on that bold statement :)Everything peaked at Castlevania: Symphony of the Night in terms of graphics and music and I will not be fielding further comment on that bold statement :)
-
Comment on Yes, you can measure software developer productivity in ~tech
SecretAgentMan Link ParentThe problem our shop had many years ago when the edict came from on high that we must now operate using Agile was that we went too by the book. Nearly all of our teams were working on established...But agile is in some sense about “find a process that works for you,” even if that seems like vacuous advice, and that might mean throwing out story points entirely.
The problem our shop had many years ago when the edict came from on high that we must now operate using Agile was that we went too by the book. Nearly all of our teams were working on established products that, occasionally, undergo major transformations that have something resembling greenfield work but are mostly driven by maintenance and small transformations.
After a year of getting pissed off at my team for constantly missing estimates and never finishing committed work for a given Sprint, I realized the problem wasn't their throughput, it was the measurement. Story Points still had value to understand complexity, but two week timeboxes were 100% arbitrary and not even in sync with our production delivery cycle (pre-CE tools, so deployments could realistically only be done about every 6 weeks).
So we just moved our delivery tracking to Kanban and kept most of the Agile ceremonies but also reduced who needed to attend them regularly (people just sitting around listening to grooming about stuff they would never work on). We stopped worrying about the end of the next sprint and instead about the next scheduled deployment date which is all that matters anyway. Made change mgmt easier and de-stressed the team from trying to make arbitrary dates to focus instead on committing to only as much work as they could realistically take on for the week. Anyone who runs light requests more or volunteers to help others complete their tasks.
You'll never get to that spot by reading the manual, but you'll know you've gotten to where you need to when it all just feels right. As a manager so long as you're staying 100% cognizant of your devs' time and staying out of their way until they start spinning their wheels, you're doing well.
-
Comment on Yes, you can measure software developer productivity in ~tech
SecretAgentMan Link ParentOne of the things I've noticed is that, even if you do this, the energy that management has for scrutinizing every little detail of software development efforts (nearly always triggered by some...I've never seen an organization measure actuals and use them to feed back into their estimation process. This step is obviously mandatory to getting good estimates, and I've literally never seen it done.
One of the things I've noticed is that, even if you do this, the energy that management has for scrutinizing every little detail of software development efforts (nearly always triggered by some upper manager who is a. annoyed by something their manager is pushing them on b. excited for some reason to try to plan out every little eventuality or c. just one of these metric-obsessed people who thinks you can predict everything with the right framework) usually peters out after too short of a period to be useful.
Basically, by the time you've sort of figured it out (or think you have), the need to report up on the estimates has diminished because you've delivered enough perceived value during that time that no one has their hackles up that your devs are just sitting around doing nothing. Or, like a golden retriever puppy, they've just moved onto the next toy.
-
Comment on Yes, you can measure software developer productivity in ~tech
SecretAgentMan Link ParentThe problem most middle managers run into is if you record an estimated amount of time for something, it's way easier for upper mgmt to hold you to unrealistic timeframes. If you obfuscate it a...drop the pretense and just estimate a duration of time.
The problem most middle managers run into is if you record an estimated amount of time for something, it's way easier for upper mgmt to hold you to unrealistic timeframes. If you obfuscate it a bit by just using Story Points/t-shirt sizing as a measure of complexity, then you have more ability to realistically work with your devs on when things can get done.
It's smoke and mirrors; in an ideal world it wouldn't need to exist. But, unless the devs also control their own paychecks and expenses, it's a reality they have to navigate. Good dev team managers should be able to shield them from most of that, but having a framework to help with it can make a difference.
-
Comment on Yes, you can measure software developer productivity in ~tech
SecretAgentMan Link ParentJira and software like it aren't inherently useful day-to-day, at least no more than any other tool or list you could create on your own would be. And in organizations past a certain size, that is...Jira and software like it aren't inherently useful day-to-day, at least no more than any other tool or list you could create on your own would be.
And in organizations past a certain size, that is still mostly true. However, as I tell my team, the value in creating a good who-what-why in your stories and making good use of comments and tagging are not in getting the work done, but in referencing it later.
Sure, you can always check your code libraries to know what you did, but knowing why you did it and who asked for it can save hours, days, sometimes even weeks of back and forth on the functional/architecture side of the house. It might not mean much to a developer doing the work, but it will if they find out they have to undo something later that ended up missing out on important business context.
Writ large: rough estimates, grouping of requests, and understanding technical complexities in some sort of reportable fashion are the only effective ways that workloads can be scheduled/managed as priorities shift over months and years.
So I always understand the gripe devs have about this, but it's the same thing you hear from salespeople and CRM: they just don't want to do the overhead for someone else, but they often lose sight of the bigger picture of why it's important. And, of course, there's ways of overdoing it. Only good managers find the balance.
-
Comment on Recommend your favorite "cozy" games, please in ~games
SecretAgentMan LinkDoes Microsoft Flight Simulator count? Once you're on autopilot and just enjoying the (mostly true-to-life) world scenery passing by below you, it can be a great way to just sort of appreciate the...Does Microsoft Flight Simulator count? Once you're on autopilot and just enjoying the (mostly true-to-life) world scenery passing by below you, it can be a great way to just sort of appreciate the world as it is.
Of course, you can get really stressful situations in it, too, but if you seek out the right type of plane and flight plan, it can be very relaxing.
-
Comment on Do advanced civilizations know we're here? in ~space
SecretAgentMan LinkI'm betting there's a reasonably high chance there are other intelligent species/societies in the universe. I'm also willing to bet only slightly less that none of those know about each other at...I'm betting there's a reasonably high chance there are other intelligent species/societies in the universe.
I'm also willing to bet only slightly less that none of those know about each other at all due to the sheer distances between them + challenges in communication/recognition.
-
Comment on Happy Baldur's Gate 3 Day! What are you going to play? in ~games
SecretAgentMan Link ParentFirst attempt was with a monk. Went ok; they certainly have some nice offensive advantages in combat, but I have no clue what the long-term outlook is with one.First attempt was with a monk. Went ok; they certainly have some nice offensive advantages in combat, but I have no clue what the long-term outlook is with one.
-
Comment on People who make smoothies, what are your favorite ingredient combinations or spices? in ~food
SecretAgentMan Link-equal parts dry oats, blueberries, and soy milk (berries can be fresh or frozen, depends on how cold you want it; substitute with strawberries if you like) -cinnamon -tablespoon of Greek yogurt...-equal parts dry oats, blueberries, and soy milk (berries can be fresh or frozen, depends on how cold you want it; substitute with strawberries if you like)
-cinnamon
-tablespoon of Greek yogurt
-tablespoon of natural peanut butter (only ingredient peanuts, nothing else)
-your choice of protein powder (I use pea protein) -
Comment on What is your favourite cutscene/cinematic in any game? in ~games
SecretAgentMan Link ParentGreat ending scene, too, with him walking through the changing seasons!Great ending scene, too, with him walking through the changing seasons!
-
Comment on What games do you most wish had a remake, or a sequel or both? in ~games
SecretAgentMan Link ParentMartian Memorandum was a masterpiece. It does feel like a series that was a product of its time in retrospect but I'm sure there would be a way to modernize it effectively.Martian Memorandum was a masterpiece. It does feel like a series that was a product of its time in retrospect but I'm sure there would be a way to modernize it effectively.
-
Comment on What games do you most wish had a remake, or a sequel or both? in ~games
SecretAgentMan Link ParentI'd like to see one for War In Middle Earth, a combination RPG and strategy game. The mechanics were interesting: you start off controlling the hobbits escaping The Shire, and nothing stops you...I'd like to see one for War In Middle Earth, a combination RPG and strategy game. The mechanics were interesting: you start off controlling the hobbits escaping The Shire, and nothing stops you from moving them anywhere on the map. But if you want to actually survive, you'll want to follow more or less the path they do in the story (so you can get Aragorn, Glorfindel, etc to help protect you).
Some of the armies are available for control right away, but others must be activated through other means which are, frankly, a bit buggy or too difficult to do within the means you're given. But it does allow for interesting alternate histories for the War of the Ring; it just really needs updates to all of the battle mechanics to make it more interesting.
Alternatively, it could be remade as two separate games - one a stealth game with small parties trying to destroy the Ring and one with just the large format battles but I like that both come into play because they both affect each other's outcomes in the original.
-
Comment on What games do you most wish had a remake, or a sequel or both? in ~games
SecretAgentMan Link ParentThis is easily my #1 choice. There is still so much that is relevant about the game, but as you say, the tech tree could probably use both a huge expansion and a reality update. The overall...This is easily my #1 choice. There is still so much that is relevant about the game, but as you say, the tech tree could probably use both a huge expansion and a reality update.
The overall mechanics of the game could stay mostly identical so long as a significant AI update commensurate with the difficulty level was also introduced. Some of the fan mods have solved for this stuff, but I've found it all to be of mixed success.
Obviously the cutscenes could be redone with modern graphics and the game map itself, too, though I really wouldn't monkey with it too much. Idk if they could track down the original voice actors to record new stuff for new tech projects.
I've seen some pieces of Justified, and it's come very highly recommended from both my wife and some friends, so I owe it a proper watch at some point.