My bottleneck is still speed of reading data from json. I wonder whether I should wait for features to be built out here or go down the path of writing a custom data reader in C++
Risk would come in the form of a weighted measure of replacement costs + loss of driving capabilities during the race. e.g. Probability of damage given a certain move/overtake, and if damages does occur taking into account the expected loss of driving capability and or cost.
Also interesting would be to see how AI would implicitly have to take into account game theory with many agents - e.g. If the car in front of 'me' has been highly aggressive in terms of defending against overtakes - what would 'my' best course of action be? Should I signal to him that his strategy is not a deterrence to me trying to overtake?
When was that xkcd posted? - couldn't not find a date.
Since machine vision is such a fast moving field, I think date is relevant for understanding the context of this xkcd. Is image recognition of a bird so inconceivable at the moment? Perhaps the joke now would be - "I'll need one researcher and one year"
This patent is pretty far reaching as it effectively just encapsulates most of the machine learning field. So does not seem enforceable (but you never know - it could take one judge to not understand).
Not sure what Google's main motive behind patenting this is but it's fun to speculate - I am hoping it's just to protect the ML field against patent trolls trying to 'weaponize' patents in the field.
Igor Feitoza, a Brazilian-born entrepreneur, left an angry message on his bank’s Facebook page this week.
“I want to see if you guys will pay my overdue bills and my employees as I can’t access my money which I . . . deposited last Friday to pay them,” he wrote following a “ridiculous” three hours spent at the bank.
Mr Feitoza is not one of the many frustrated Royal Bank of Scotland customers hit by the latest technology failure this week. He is a client of Commonwealth Bank of Australia, the country’s biggest bank by market capitalisation, which had its reputation as one of the most digitally advanced lenders tarnished by an outage in its payment and online systems late last week.
As well as provoking outbursts on social media, the technology glitches at RBS, which caused some 600,000 payments and direct debits to go missing, and CBA underline how the world’s biggest banks are often failing to get to grips with the growing demands being placed on their IT platforms.
Banks spent about $188bn on IT last year and that figure is expected to grow at close to 5 per cent a year, taking it above $200bn by next year, according to Celent, the research company.
Many banks, such as RBS, are plagued by computer systems that have been built up over several decades through acquisitions and new product launches to form a costly and complex patchwork of systems. “A lot of these programmes get three-quarters done,” says a senior technology executive at one of the largest US banks. “It’s one of these technical debt problems that builds up over the course of time. If you have one thing happen it might cause the whole thing to fall down.”
The cost of maintaining these often ageing and unwieldy systems eats up three-quarters of banks’ IT spending, according to Celent. That leaves only a quarter to spend on innovations to keep up with the rapidly emerging threat from the many technology groups and start-ups trying to steal market share in areas such as payments.
“For a sector that spends significantly more on technology than most other sectors in the world, it is the least innovative, so there is a paradox here,” says Bill Michael, head of financial services in Europe at KPMG.
As many banks struggle in the post-financial crisis environment to generate returns above their cost of capital, these spiralling costs and inefficiencies are becoming increasingly unacceptable to managers and shareholders.
Bank-IT-spend-chart
“The banks are all putting their hands up and saying: ‘Help, I no longer want to do this alone any more — can someone else come and help, whether that be a cloud services provider or a partnership with a start-up’,” says Mr Michael.
RBS, which last year paid a record fine to regulators for a bigger systems outage in 2012, hoped to solve its problems by replacing its core processing engine at a cost of £750m. But in a recent interview, its chief executive Ross McEwan conceded there was still a big job to reduce the number of systems and applications at RBS from more than 3,000.
Andrea Orcel, global head of UBS’s investment bank, says: “The challenge for most banks is that they are not technologists ... As technology continues to evolve at a fast pace, becoming ever more critical to their business, they are having to navigate a space that is both highly complex, and does not play to their core competencies.”
Asian banks spend more than Europeans or US rivals on IT; their spending is growing faster; and more of their IT dollars are going on new projects than on maintenance.
However, they are hardly glitch-free. In January, a problem in the system linking accounts at Industrial and Commercial Bank of China, the country’s largest lender, with securities brokerages, disrupted Rmb4.9bn in fund transfers, affecting nearly 55,000 customers at 90 brokerages.
Bank-IT-spend-chart
As regulators make ever growing demands on banks to provide them with vast amounts of data for everything from stress tests to anti-money laundering checks, banks are racing to keep their systems up to speed.
Deutsche Bank insiders blamed its failure in this year’s US stress test on years of under-investment in IT that made it unable to meet US regulators’ demands.
Concern is growing about cyber security after a string of high profile hacking attacks, such as last year’s theft of data on 76m customers from computer systems at JPMorgan Chase.
Executives say this focus on cyber security is a catalyst for change, pushing banks to simplify and upgrade their IT systems. “Making something secure requires it to be consistent and clean and up to date and well managed,” says the US tech executive.
James O’Neill, senior analyst at Celent, predicts that within a decade most big banks will have switched from using costly mainframe computers for overnight processing of customer data to using much more flexible cloud-based services.
If nothing else, the shortage of developers trained in the Cobol programming language that drives most bank mainframes will force them to make the switch. Australia’s CBA has moved to a cloud-based system, while Deutsche recently outsourced many of its applications to a cloud provider.
Bank-IT-spend-chart
“The story about legacy systems impeding innovation is a bit oversold,” says Mr O’Neill. “I’m not saying they will go on for ever, but they have done a pretty decent job of supplying new products and services.”
Like many sectors, however, banks are realising the need to harness the power of “big data” to offer better digital services to their customers.
Mr Michael at KPMG says his banking clients know the stakes are high. “They have seen what technology has done to music and home shopping and so far they have been relatively slow to adapt — but the longer they wait the harder it will get.”
The reason 4chan works is that 4chan itself is very like-minded. Secret attracted too broad an audience and scared off the quality content creators as bad content started rising. Divergence of 'like-mindedness' among different social groups was just too large for one feed.
In my area Secret had a similar story - started out being popular with people in their 20+'s, putting out some insightful/funny content (perhaps with some light local context). Content slowly started spiraling downward toward very specific sort of circle-jerked inside jokes (e.g. jokes about some high-schools, fraternities and even specific kids at larger schools).
Anyway it is a shame because there were people out there put out genuinely good content.
The experience is similar with the app 'Fling' (Send snapchat-like picture to 50 random people) 70% of the content is guys lying in bed with 'lol hi any gurls?'.
Perhaps a way to remedy this would be to cluster on 'like-mindedness'. Then have content show more according to whatever cluster you are in. Constant feedback on what you like and what you put out.
Some good ML + NLP based filtering could give the impression at least that many like-minded people are using the app and inspire people to put out content. If you don't allow users to control the content they follow (Facebook / twitter / vine / youtube / snapchat - 'supervised') you have to direct the right kind of content to users in an 'unsupervised' way.