I once calculated that such a system, if applied in Canada, would require something like a 36% flat income tax if it were entirely funded that way, and it would have a break-even point at about $65k annual income (below that, you receive more than you put in). I expect the situation in the US to be similar.
Part of that cost could be offset by reducing or eliminating existing welfare and social security programs, including food stamps, employment insurance, and subsidized housing. It would also let you eliminate minimum wage, since you no longer have to be concerned about wages covering basic living expenses.
There are a lot of questions about how this would affect an economy. Would it cause massive inflation? Unemployment? Substance abuse? Or would it be the most effective social program ever? I have no idea. But I think it's worth a try. My hunch is that you will see all of those negative effects increase somewhat (inflation, unemployment, substance abuse) or in some people, but that the positive effects will greatly outweigh the negatives.
You needn't implement it all at once. You could apply it first to a random sample, say 1% of the population. Those people receive basic income and pay the tax required to fund it. They would also become ineligible for the social programs that would be removed if you rolled it out nationally (the details of that part may be tricky). Only roll it out to the whole population if the results are positive after a few years.
It's hard to say how things will balance out. That's why I'm in favour of some sort of systematic test of the concept.
The idea that people can allocate resources for themselves better than somebody else can do for them has an interesting tie-in to today's debate about what policies we should pursue to stimulate economic growth. The basic Keynesian proscription for reviving a moribund economy is to put idle resources to work through more government investment, and for many years economists have tried to figure out which investments offer the best bang for the buck by measuring fiscal multipliers, etc. But this approach assumes that a dollar of output is a dollar of output, when in reality—if we're doing economics, and not mere accounting—people often value a dollar's worth of one good very differently from a dollar's worth of another. (And a disparity between accounting value and economic value is likelier to exist when investment decisions are made by politicians, at a remove from the mainstream of economic activity.)
This suggests we should try to allocate stimulus spending in a way that mirrors as closely as possible the distribution of resources we'd get if our economy were chugging along at a good clip, with all our people and capital doing the things the market would arrange if were working properly. Greg Mankiw, the chairman of the Harvard economics department, wrote a paper with his colleague Matthew Weinzierl [0] a couple of years ago that tries to get at what that would look like. Their simple model suggests, unsurprisingly, that traditional make-work government investment programs should be a last resort: getting the same amount of money into the hands of investors and consumers who can allocate it more effectively might be, well, more effective.
0. http://scholar.harvard.edu/mankiw/files/exploration_of_optim... [PDF]
Everyone (well, every citizen between, say, 18 and 67, when Social Security kicks in) gets a guaranteed flat income from the government. Then raise income taxes accordingly such that the average middle class family more or less breaks even.
What is my benefit for such a compromise?
Or is the argument that the social good created by BI would outweigh such cases?
Hopefully, with the success of such charities in developing countries we will see increasing support for such systems here at home, allowing people to escape the vicious unemployment trap of western economies.
(Edit) Additionally, the implementation of basic income at this point would be extremely effective, as it would leverage the massive revolution in freely available online education (e.g. Khan Academy, EdX, Udacity, Coursera) and cooperative workspaces (e.g. hackerspaces). This would effectively transform the entire country into a startup incubator, giving those with the desire and talent to build new companies the resources to do so immediately, without going through the excruciatingly slow, inefficient and oftentimes counterproductive process of formal education.