Tuesday, July 5. 2016
The Wisdom of Crowds showed us that a group of people were wiser than a group of experts - and then came the caveats. Firstly, it was shown that all the Wise people hae to decide independently of each other.
Now, it seems that small is better - Santa Fe Institute
Sounds suspiciously like a group of experts again, though there is the random selection requirement again. It also depends what the question is:
Where previous research on collective intelligence deals mainly with decisions of how much or how many, the current study applies to this-or-that decisions under a majority vote. The researchers mathematically modeled group accuracy under different group sizes and combinations of task difficulties. They found that in situations similar to a real world expert panel, where group members encounter a combination of mostly easy tasks peppered with more difficult ones, small groups proved more accurate than larger ones. This effect is independent of other influences on group accuracy, such as following an opinion leader or having group discussions before voting.
What about voting as a means of determining the majority opinion of a populace?
"These results, of course, do not mean that we should abandon large scale referendums like Brexit and national elections,” Galesic adds. “Choices between different policies and candidates often do not have a 'right' and a 'wrong' answer: different people simply prefer different things, and the outcomes of these decisions are complex, with a spectrum of consequences. It is important to account for everyone's opinion about the general direction in which they want their country to go -- including underrepresented groups.
Not sure where this leaves us practically though - for some problems it's better to have smaller groups, others larger, depending on the question.
Sunday, July 3. 2016
The first fatal crash of a "self driving car" is pushing a lot of questions to the surface, on a whole lot of levels.
Is the Technology up to it yet?
Firstly, its is clear now that the technology is not yet ready for wide scale deployment - a camera probably should not be the prime mode of sensing (night and difficult lighting conditions, lens obstructed) and the current radar is clearly sub-optimal - there are a lot of potential obstacles below car roof height.
Was it tested enough?
These cars have done many millions of miles, and are said to be less risky than driving you own car (it has twice the miles/death ratio of average motoring - but the well heeled Tesla driving demographic is far from average). Another key question is what sort of miles? Has it been pushed hard, beyond the envelope, by test drivers. These are not "far-beyond the edge" driving conditions. It almost smacks of software industry culture - push a Beta product out there, let the customers find the bugs. But bugs in heavy, powerful, fast mechatronic devices can kill.
It was billed last year as the "arrival of your autopilot". Problem is, some people believed it. Fortunately, not too many people can afford them, and as mentioned above most of those who bought it are well heeled - less envelope pushers as a % of drivers. But there are some, posting videos of his experiments with hands free driving on YouTube (the driver concerned was one such)
The inevitable outcome - Regulation - is probably a Good Thing right now
As the WSJ explains, despite misgivings regulators were persuaded to stay their hand:
Auto-safety regulators, meanwhile, were relatively silent on the technology even though many experts viewed Tesla’s program as the most aggressive self-driving system on U.S. roads. The National Highway Traffic Safety Administration, embroiled in managing a sharp increase in safety recalls, including tens of millions of rupture-prone air bags, lacks authority to approve or disapprove of the advanced technology or meaningfully slow its deployment.
Now they will. And despite AI-Car supporters' cries of "men walking with red flags in front of cars" it is a necessary, and in the medum run a good step for the self driving car industry to get it a lot more right first.
What will kill the AI-Car industry stone dead is if it kills a few more people.
Friday, July 1. 2016
As is becoming usual, most pundits got the British EU Referendum (Brexit) winner wrong (to be fair, the two sides were so close running that most were within statistical margins of errors, just that most picked the wrong side as the winner - and that makes all the difference of course, just ask the speculators who were caught). But one company, TNS, got it right. Being datawonks we were fascinated about how they did it - summary below from article in El Reg:
1. Balance out the Politically Engaged (mainly Remainers) to correct for shy people
We asked respondents about their likelihood of voting in the next General Election and used this (along with some demographic information) to model turnout using data we collected at the 2015 General Election. We compensated for this imbalance by weighting the turnout level of our sample down to a more realistic level; decreasing the number of politically engaged individuals and giving us a more representative sample.
2. Rebalance using known demographic drivers (education, class etc)
3. Remove Confirmation bias
There was also added risk due to the fact that 16 per cent of registered voters were still undecided and we were unsure as to whether they would vote. Our understanding is that some polling companies tried to take this into account by allocating a higher proportion of the undecided voters to Remain; this also seems to have been factored in by the betting market as it consistently showed a higher probability of Remain than the opinion polls. [TNS did not do this]
I can't help but hypothesize, given how everyone was clustering around the same broad numbers, that the mental model that meant they did not assume the "Undecideds" would vote one way or the other was what tipped it.
Update - article here showing endemic errors in the poling methods meant that Leave as always in the lead, but polls couldn't se it
Friday, June 24. 2016
Today, the UK voted to leave the European Union (EU), whatever that may mean.
Whatever you think of the outcome (this are not really a political blog), we are officially in Uncharted Seas, in Interesting Times - History is beginning again
From a technology point of view there are some interesting questions, for example about the UK's adoption of EU data rules, or whether the UK "Tech" sector is better of relocating to Berlin or Dublin or somesuch, or whatever.
We have been going for 10 years this year, and have had to do some quite interesting predictive work over the last 10 years, from the market opportunity of niche products future of entire industries, but we would never have predicted this until a few weeks ago when it was clear social media sentiment was rapidly shifting
Change is a constant.....the next 10 promise to be just as interesting.
Monday, June 13. 2016
It is hard to see exactly what synergies the deal brings - picture above courtesy Matt Zeitlin
Microsoft has bought Linked In, the question top of mind to us is "why - and why pay so much?".
They have paid about 50% over the odds (Linked In shares were c $130, now are c $195 on MSFT bid of $196), that is a considerable premium.
It would seem to be a bet on the "Future of Work" - Satya Nadella (MSFT CEO) says that:
Quite how this translates practically is hard to see, as the diagram shows above synergies are not exactly obvious so its an accretion play. But the business cases trotted out are speculative and not particulalry compelling:
- Microsoft Office combined with LinkedIn's network so Microsoft can point to a specialized expert through LinkedIn
This is not the stuff of a 50% level of valuation premium, it's fluff for diverting tech journos. Yet Nadella is no fool, so what is really in play here? Thinking laterally, it gives MSFT access to a lot more data about YOU!:
- Access to the social graphs and details about a lot of working people globally, i.e. the list of nearly every customer, and insight into many companies that Microsoft has or wants - a CRM system wet-dream
Now that is the sort of thing that has got real value, and the price then starts makes sense - deter others from entering any bidding war.
Of course, it could not be be, as one wag on Twitter suggested, MSFT's attempt to "consolidate its dominance over the most joyless aspects of your computing life" - though if we follow Lewinsky's Law, that the most dull tech is the most profitable, it also explains the valuation
Friday, June 3. 2016
It gets, er, better - from troncInc's own press release:
“tronc pools the company’s leading media brands and leverages innovative technology to deliver personalized and interactive experiences to its 60m monthly users,”
And then there is troncX, an “online curation and monetization engine” which utilizes artificial intelligence technology “to accelerate digital growth”.
Oddly enough, the marketplace has not been wowed, in fact some have been uncouth enough to even suggest that the brand name was badly troncated, or even that the Branding Consultants were tronc (OK, OK, nearly everyone is hooting with laughter and taking the piss)
But all is not lost - firstly, its an excellent Wildean Strategy and lets not ignore that they have all those right-on-the-money buzzwords - Monetization, premium, AI, digital, accelerate - and it starts with a small letter to boot. Not a bad start for Unicorn bingo, but Broadstuff analysis shows a serious flaw - surely, surely to hit max points they should have gone for that double "oo" thing?
You know it makes sense....besides, what (more) could go wrong?
Friday, May 27. 2016
Godwin's Law of Online Discourse - that as an online discussion continues, the probability of a comparison to Hitler or to Nazis approaches 1 - has been proven.
From Mr Godwin himself:
On the law itself, he notes that:
That hasn't gone so well it seems...he also noted the propensity of London mayors to invoke it nowadays:
Tuesday, May 17. 2016
BBC is being required to reduce its spend by our current Government, and one of the decisions is to remove 11,000 food recipes built up over 15 years or so:
A number of interesting lessons from this about electronic vs. paper media:
- Online data is not "permanent" - it can be removed at the whim of economic (or in this case, more likely political) changes in the winds. Once gone, its not clear it can be returned
This being the UK, there is already a campaign and a petition to keep the recipes (and perish the thought the BBC knows this would have happened)
The recipes are part of a broader picture of a larger attack on the BBC's taxpayer funded free content model, which is free for citizens to use and hurts the business models of commercial online businesses.
The BBC represents a real alternative to Ad-funded online content, essentally its a tax funded high quality content model, and has shown itself to be largely superior to the Ad-funded model in the UK and thus represents an existential threat to Ad funded approaches, hence the increasing attacks on its online assets.
Thursday, April 21. 2016
This shouldn't surprise readers of this blog, but there is a nice analysis of the problems Unicorns have with the funding rounds with guaranteed payouts to certain funders (ratchet, prefs etc) on "abovethecrowd" blog - "dirty deals" made by "shark investors"
One is that they “unpack” or “explode” at some point in the future. You can no longer simply look at the cap table and estimate your return. Once you have accepted a dirty offering, the payout at each potential future valuation requires a complex analysis, where the return for the Shark is calculated first, and then the remains are shared by everyone else. The second reason they are a massive problem is that their complexity will render future financings all but impossible.
As the blog points out, later-stage investors may be tempted to become Sharks themselves and start including "dirty deal" into their own term sheets. They will need to embrace a Shark culture, and be comfortable knowing that they are adverse to and in conflict with the founders, employees, and other investors on the capitalization chart. This is filed under "No Shit, Shylock" in Broadstuff Towers...
Troubled waters lie ahead for many of the Unicorns.....clearly these deals have inflated the Unibubble, but it looks like they will also be a major reason it bursts.
(Update - this article by Sarah Lacy notes the author of the above is an early stage VC, and thus one of those being scraped by the "dirty deals" I assume. Now while one feels a tad sorry for him, he must have known (i) the VC game, (ii) the potential burn rates of these businesses trying to colonise industries with neglibible costs of entry and (iii) that they have been in the Bubbletime for c 3 years - heck, we did, this is not hard!
Thursday, March 24. 2016
Now who would have predicted this - Sarah Perez at TC:
The clue is Sarah's line "Given that this is the Internet" - it was an 800lb Gorilla "Given"
Cue a million human internet trolls thinking today "hey, with one of those I am unstoppable 24 x 7"
The problem AI has is that it potentially promises everything and the Hypesters run with that as fact, but in truth today's early AI systems are more like intelligences with Savants syndrome - there's one thing they are taught to do very well, but are completely clueless at anything else. Deep Mind can beat Lee Se Dol at Go, but it can't do any of the other things Lee can do (and it burns 50,000 times more energy in not doing it....). Tay can "learn" (aka parrot) what people tell it but cannot distinguish between deep truths and inflammatory statements.
Oh - one last thing - Tay is modelled on a teenage girl. Sheeesh.
To be fair this is not Tay's fault per se, that you have to lay at the hands of its "parents" - how did they not see this one coming (and, when it started, who was watching its first faltering steps on its first day out playing in the traffic?). Anyone who has been around the internet for more than a few months (or even days - Boaty McBoatface anyone) would have seen that this was a not improbable risk.
PS There is a deeper story here, about the reliance on machine intelligences before they are ready, and the damage they can do without anyone knowing. The thing about complex systems is it's often very hard to find out when they are not working properly, malfunction is often not obvious - so it is necessary to watch them very closely, not let them off to roam on the Interwebz like Tay.
More Broad Stuff
Poll of the Week
Will Augmented reality just be a flash in the pan?
Creative Commons Licence
Original content in this work is licensed under a Creative Commons License